BiomarkersofAlzheimer-AssociatedEndosomalDysfunction

JessicaNeufeld

Submittedinpartialfulfillmentofthe

requirementsforthedegreeofDoctorofPhilosophy

intheGraduateSchoolofArtsandSciences

COLUMBIAUNIVERSITY

2018

©2018

JessicaNeufeld

Allrightsreserved

ABSTRACT

BiomarkersofAlzheimer-AssociatedEndosomalDysfunction

JessicaNeufeld

EndosomaldysfunctionhasbeenmechanisticallylinkedtoAlzheimer’sDisease(AD).Todate,no

invivobiomarkersforthiscellulardeficitexist.Yetsuchbiomarkersarerequiredfordetermining

its prevalence in AD and tracking its time course—both in disease progression and potential

clinicaltrials.Withthisgoalinmind,wemadeuseofanassortmentofmousemodelsbearing

AD-relatedendosomaltraffickingdefectsthroughselectivedeletionofretomercoreproteins.

WecollectedCSFandbrainexosomesfromtheseretromer-deficientmodelsandperformeda

battery of molecular inquiries which included lipidomic and proteomic screens, as well as

hypothesis-drivenbiochemistry.Theresultsofthiscomprehensiveinvestigationincludethefirst

characterization of themurine CSF lipidome and the deepest characterization to date of the

murineCSFproteome.

Herein,we report that VPS26a haploinsufficiency in the brain imparts no detectable protein

changesintheCSFasmeasuredbylabeledLC-MS/MSatthreemonthsofage.Thisdeficitdoes,

however,causeareliablereductionofCSFsphingomyelind18:1/18:1,whichisexacerbatedby

age,extendingtoothersphingomyelinsandotherlipidclassesincludingdihydrosphingomyelins

andmonohexosylceramides.

CompleteknockoutofitsparalogVPS26bpromotesanenrichmentofBACE1-cleavedAPPCTFs

(b-CTFs)inbrain-derivedexosomesandmayalterexosomalbiogenicpathways.Similartrends

wereseeninaneuronal-specificknockout(viaCamk2a-Crerecombinase)ofretromer’slinchpin,

VPS35.

Most importantly, anunbiasedproteomic screenofCSF collected frommicewitha selective

knockoutofVPS35 in forebrainneurons(engineeredusingtheCamk2a system)uncovereda

totalof71hits(52parametricand19nonparametric)fromthe1505proteinsdetected.Pathway

analysis and follow-up studies identified two distinct molecular categories with previously

establishedrelevancetoAD:BACE1substratesandMAPT(morecommonlyreferredtoastau).

Wereportthat,bothinvivoandinvitro,neuronal-selectiveknockoutofVPS35causesincreased

secretionoftheN-terminalfragments(NTFs)ofBACE1substratesAPLP1andCHL1aswellastotal

tau, and importantly, that these events occur independent of cell death. Further, we find

evidenceofconvergenceofthesepathwaysinbothmouseandhumanCSF.However,asthese

BACE1 substrates likely accumulate in plaques, we propose CSF total tau as a biomarker of

endosomaldysfunctionwithutilityovertheentirecourseofADprogression.

We have identified and validated a series of in vivo biomarkers that are reflective of AD-

associatedendosomaldysfunction.Whileclearlysensitivetothiscellularpathology,futurework

isrequiredtodeterminetheirspecificity.Additionally,follow-upstudiesarerequiredtoshow

that interventions which rescue endosomal dysfunction affect this molecular profile. The

identifiedbiomarkersholdgreatpromiseforearlydetectionofendosomaldysfunctioninADand

fortrackingitscourse,duringthediseaseprogressionandforclinicaltrials. Furthermore,the

unexpectedbutvalidatedfinding,showingthatincreasedCSFtauisreflectiveofAD-associated

endosomaldysfunction,suggeststhatendosomaldysfunctionisauniversaldeficitsharedamong

ADpatientsinitsearlieststagesofdisease.

i

TableofContents

ListofFigures………….………………………………………………………………………………………………………….…….iiiListofTables.……………………………………………………………………………………………………………………..…….iv

1 Chapter1:Introduction 11.1 Alzheimer’sDisease 1

1.1.1 Epidemiology..........................................................................................................................11.1.2 Riskfactors.............................................................................................................................21.1.3 Pathologicalhallmarks...........................................................................................................41.1.4 Diagnosticchallenges.............................................................................................................51.1.5 Theamyloidhypothesis.........................................................................................................61.1.6 Theendosomeasanemergingtherapeutictarget.............................................................11

1.2 Theretromerassembly 141.2.1 Retromerasamasterconductorofendosomaltrafficking................................................141.2.2 RetromerinAD....................................................................................................................181.2.3 RetromerdeficiencyasamodelofAD-relatedendosomaldysfunction............................19

1.3 Thesearchforbiomarkersofendosomaldysfunction 20

2 Chapter2:CSFLipidomics 252.1 Introduction 25

2.1.1 Lipidsandtheendosome.....................................................................................................252.1.2 LipiddysregulationinAD.....................................................................................................252.1.3 Massspectrometryoflipids................................................................................................262.1.4 CSFlipidomicsstrategy........................................................................................................262.1.5 Choiceoftheretromer-deficientmodel.............................................................................27

2.2 Results 272.2.1 CollectionofhighqualityCSF..............................................................................................272.2.2 ThemurineCSFlipidomecloselyresemblesthatofhumans.............................................292.2.3 LipidomicscreenrevealssphingomyelinchangesintheCSFofVPS26a-haploinsufficient

mice.....................................................................................................................................312.2.4 ReductionsinVPS26a+/-CSFsphingomyelinsaremorepronouncedwithageandextend

tootherlipidgroups...........................................................................................................362.2.5 SphingomyelindysregulationintheVPS26a+/-brainisnon-neuronal...............................38

2.3 Discussion 40

3 Chapter3:BrainExosomes 423.1 Introduction 42

3.1.1 Exosomebiology..................................................................................................................423.1.2 Exosomesasreportersofendosomaldysfunction.............................................................423.1.3 Brainexosomebiomarkerstrategy.....................................................................................43

3.2 Results 443.2.1 Isolationofbrain-derivedexosomesfromwildtypemice:validationoftechnique..........443.2.2 Retromercoreproteinsarepresentinbrainexosomefractions.......................................463.2.3 Brainexosomesderivedfromretromer-deficientmicehavereducedretromerprotein

levels...................................................................................................................................473.2.4 Brainexosomesderivedfromretromerknockoutmiceareenrichedinb-CTFs...............49

ii

3.2.5 Brainexosomesfromretromerknockoutmicehavedecreasedlevelsofexosomalmarkersalixandflotillin-1..................................................................................................52

3.3 Discussion 54

4 Chapter4:CSFProteomics 564.1 Introduction 56

4.1.1 Endosomaldysfunctionandthesecretedproteome..........................................................564.1.2 CSFProteomicsbiomarkerstrategy....................................................................................574.2.1 CSFfromVPS26a+/-micerevealsnoapparentproteinalterationsasmeasuredviaLC/MS-

MS.......................................................................................................................................584.2.2 MassspectrometryrevealsenrichmentofBACE1substrateN-terminalfragmentsand

totaltauinCSFofVPS35Camk2aknockoutmice.............................................................634.2.3 BACE1SubstrateNTFsandTotalTauareconfirmedashitsinanewcohortofVPS35

Camk2aknockoutmice......................................................................................................714.2.4 TranslationtohumanspositionsCSFtotaltauasabiomarkerofendosomaldysfunction

inAD....................................................................................................................................754.3 Discussion 79

4.3.1 VPS26a+/-CSFProteomics....................................................................................................804.3.2 VPS35Camk2aCSFProteomics...........................................................................................80

5. Chapter5:Conclusionsandfuturedirections 865.1 Overview 865.2 CSFlipidomics 865.3 Brainexosomestudies 875.4 CSFproteomics 89

6. MaterialsandMethods 93

7. References 102

iii

ListofFigures

Chapter1Figure1.1ThegeneticsofAD. ....................................................................................................... 3Figure1.2ThepathologicalhallmarksofAD. ............................................................................... 5Figure1.3Theprocessingpathwaysofamyloidprecursorprotein(APP). .................................. 7Figure1.4Theamyloidcascadehypothesis. ................................................................................. 8Figure1.5EndosomaltraffickinginAD. ...................................................................................... 12Figure1.6Theretromerassembly. .............................................................................................. 15Figure1.7ThesearchforbiomarkersofAD-relatedendosomaldysfunction. .......................... 21Figure1.8Mousemodelsofretromer-dependentendosomaldysfunction. ............................ 24Figure2.1CollectionofhighqualityCSF. .................................................................................... 28Figure2.2ComparisonofthemouseandhumanCSFlipidome. ............................................... 30Figure2.3VPS26a+/-CSFlipidomicsstudyoverview. .................................................................. 32Figure2.4LipidomicscreenrevealsalteredsphingomyelinprofileinVPS26a+/-CSF. .............. 33Figure2.5Follow-upstudyinnewcohortconfirmsdecreaseofSMd18:1/18:1inVPS26a+/-

CSF. ........................................................................................................................................ 35Figure2.6AgeexacerbatessphingomyelinprofileinVPS26a+/-CSFandextendseffectsto

otherlipidclasses. ................................................................................................................ 37Figure2.7SphingomyelinalterationsinVPS26a-haploinsufficientmicederivefromnon-

neuronalorigins. .................................................................................................................. 39Figure3.1Isolationofhighqualityexosomesfromadultmurinebrain. .................................. 45Figure3.2Retromercargorecognitioncoreproteinsarepresentinbrainexosomefractions.47Figure3.3Murinebrainexosomesreflectretromerdeficiencyinthebrain. ........................... 48Figure3.4Retromerdeficiencyinvivopromotesb-CTFaccumulationinexosomes. ............... 51Figure3.5BrainexosomesfromVPS26bknockoutmicecontainreducedlevelsofcanonical

exosomemarkers ................................................................................................................. 52Figure3.6DescriptiveelectronmicroscopyrevealsenrichmentofsmallervesiclesinVPS26b-/-

brainexosomes. ................................................................................................................... 53Figure4.1OverviewoflabeledLC-MS/MSapproachforproteomicanalysisofVPS26a+/-CSF.

............................................................................................................................................... 59Figure4.2Qualityassessmentrevealsnormallydistributedproteinquantitationforall

samples. ................................................................................................................................ 60Figure4.3GlobalCSFprofileinVPS26a+/-micelargelyresemblesthatoflittermatecontrols.

............................................................................................................................................... 61Figure4.4PAFAH1B3isunchangedinVPS26a+/-CSF. ................................................................. 63Figure4.5OverviewoflabeledLC-MS/MSapproachforproteomicanalysisofVPS35Camk2α

knockoutCSF. ....................................................................................................................... 64Figure4.6ComparisonwithpriormurineCSFproteomicstudies. ............................................ 65Figure4.7Qualityassessmentrevealshighlevelsofcorrelationbetweenreplicates. ............. 66Figure4.8BACE1substrateNTFsareelevatedinCSFfromVPS35Camk2αknockoutmice. ... 69Figure4.9NonparametricanalysisrevealspresenceofMapt(tau)inonlyretromerknockout

CSF. ........................................................................................................................................ 70

iv

Figure4.10IngenuityPathwayAnalysis(IPA)implicatesAD-relatedproteinsasupstreamregulatorsinVPS35knockoutCSF. ...................................................................................... 71

Figure4.11ElevationofBACE1substrateNTFsinVPS35knockoutsisconfirmedinvivoandinvitro. ...................................................................................................................................... 72

Figure4.12TotaltauissignificantlyelevatedinVPS35Camk2αknockoutCSF. ....................... 74Figure4.13TotaltauisenrichedinmediafromVPS35-/-neuroncultures. ............................. 75Figure4.14NTFsofBACE1substratesarecorrelatedinmouseandhumanCSF. ..................... 76Figure4.15APLP1NTFsarehighlycorrelatedwithAβ42andtotaltauinhumanCSF. ............ 77Figure4.16CSFLevelsofAPLP1NTFsarelowerinADpatients. ............................................... 79

ListofTables

Table4.1ParametricHitsofVPS35Camk2aCSFProteomics. ................................................... 66

v

Acknowledgments

Thescientificpursuitisasubstantialundertaking,asthepathfrombeginningtoendisrarely

straightorentirelyprescribed.Accordingly,itrequirespassion,dedication,andoftenasizeable

supportnetworktofulfillone’sinitialobjective.Iamblessedtolayclaimtosuchanextensive

andunwaveringcast,withoutwhich,thisjourneywouldnothavebeenpossible.

Ifirstwishtothankmyfellowlabmateswhohavesharedtheirknowledge,expertise,andbench

spacewithmeoverthepastfewyears.Inparticular,IwishtoacknowledgeSabrinaSimoes,

whoassistedwiththeelectronmicroscopyforthisprojectandwhohastaughtmemoreabout

scienceandlifethanIevercouldhaveasked.ImustalsothankAndreaUrban,whose

exceptionaladministrativesupportandholidaygoodieskeptallturbulenceatbay.Finally,I

wouldliketoacknowledgeMilankumarKothiyaforhisoutstandingtechnicalassistanceinthe

finalphasesofthisproject.

ToourcontributingcollaboratorsEmilyChen,EstelaAreaGomez,RobinChan,YimengXu,Larry

Honig,DavidKulick,LaurelProvencher,andPurvishPatel,Iextendmyutmostgratitude.Iwould

alsoliketoacknowledgeBeverlySheltonandXiaolanShen,whoprovidedexperttrainingin

murineCSFcollectiontechniques,thecornerstoneofthisproject.

IamindebtedtothePathologyDepartmentforgrantingmethisexceptionaltraining

opportunity,andtoZaiaSivoforherunendingadvocacy.Ialsowishtoacknowledgemythesis

vi

committeemembers,CarolTroy,ClarissaWaites,GilDiPaolo,fortheirexceptionaladvisement

throughoutthecourseofthisproject,andtoAndrewTeichforhiswillingnesstoserveas

additionalcounselforthedefense.

Iwishtothankmyfamilywhohasalwaysencouragedmetoshootforthestars,andsupported

mydedicationingettingthere.Andinparticular,mybrotherJason,whoseweeklyphonecalls

kepteverythinginperspective.

Myworkhasbeenfurtherblessedbythesupportofcountlessfriends.Iamparticularly

indebtedtoNoopurKhobrekarandAidanQuinnforthegiftsoftheirtimeandscientific

wisdom,andtoLyudaKovalchukeandYelenaAkhtiorskaya,towhomtheword‘friendship’

knowsnobounds.IalsowishtothankFabriceMercier,wholearnedmoreaboutretromerin

thepastthreeyearsthanheeverintended,andwhoseloveandgheegotmethroughallof

life’sgreatestchallenges.

Finally,Iwouldliketothankmythesisadvisor,ScottSmall,forhisinvaluablementorshipover

thecourseofthisproject.Scotthasalwaysprovidedboththetoolsandthecounselnecessary

tosucceed.Andimportantly,hisconfidenceinmyabilitytodosohasneverwavered.Scotthas

shownmewhatitmeanstobebotharigorousscientistandadiligentscholar,andhisunending

supportalongbothtrajectorieshasaffordedmetremendousgrowth.

1

1 Chapter1:Introduction1.1 Alzheimer’sDisease

1.1.1 Epidemiology

Alzheimer’sdisease (AD) isadevastatingneurodegenerativedisease,affectingmore than5.5

millionAmericans.itiscurrentlytheleadingcauseofdementia(Hebert,Weuveetal.2013)and

thesixthleadingcauseofdeath(KochanekKD2016).ADdiagnosishingesuponaconvergenceof

pathologicalandcognitivesymptoms;yet,heterogeniccomplexitiesposechallengetodiagnostic

accuracy and therapeutic intervention. Nevertheless, AD is broadly categorized into two

subgroupsbasedonageofonset:early (EOAD)and lateonset (LOAD).The formercomprises

thosebelowtheageof65,whileLOADincludespatients65andolder.

MostADcasesappearsporadicallywithoutanyclearinheritancepattern.However,dominantly

inheritedpatternshavebeenreported,andareestimatedtorepresentbetween1%and5%of

all cases (Bird 1993). This dominantly inherited AD often presents at an earlier age and is

thereforeoftentermedearlyonsetfamilialAD(eFADorEOFAD).

Nearlytwo-thirdsofthoselivingwithADarewomen(Hebert,Weuveetal.2013).Thereasonfor

this sex-based discrepancy remains unclear, although social (e.g. inequitable educational

attainment for thecurrentADpopulation)andbiologicaldistinctions (e.g. higherassociation

withtheAPOE-4genotype)havebeenproposedascontributingfactors (Carter,Resnicketal.

2012,Altmann,Tianetal.2014).

2

1.1.2 Riskfactors

ThemajorityofADsufferers(96%)are65orolder(Hebert,Weuveetal.2013).Infact,oneinten

adults over 65hasAD and theprevalence increaseswith age (up to 32%of adults over 85),

renderingagethebiggestriskfactorforthedisease(Hebert,Weuveetal.2013).

Asecondriskfactorisgeneticpredisposition,involvingtwotypesofassociations:deterministic

genes and risk genes. Deterministic genes are dominantly inherited and disease-causing.

MutationsinAPP,PSEN1,andPSEN1comprisethelargemajorityofthesecases(Bird1993);more

recently,rarebutcausalmutationsinSORL1havealsobeenreported(Pottier,Hannequinetal.

2012,Vardarajan,Zhangetal.2015,Verheijen,VandenBosscheetal.2016).

Riskgenes,ontheotherhand,increaseone’slikelihoodtodevelopthedisease,asdetermined

bygenomewideassociationstudies(GWAS).Suchgenes includeAPOE,whichiscurrentlythe

largestgeneticriskfactorforsporadicAD.HeterozygosityfortheAPOE4alleleincreasesone’s

risk for developing AD roughly 3-fold, while homozygosity increases the risk 12- to 15-fold

(Michaelson2014,WuandZhao2016).Interestingly,therareAPOE2alleleappearstolowerthis

risk(WuandZhao2016).AgrowingnumberofothergeneslinkedtoADhavebeenidentified

throughgeneticsstudiesinrecentyears(Figure1.1)(Guerreiro,Brasetal.2013,KarchandGoate

2015),thoughnoneconferashighariskfordiseasedevelopmentasAPOE4.

3

Figure1.1ThegeneticsofAD.

AgrowingnumberofgeneshavebeenlinkedtoAD,includingbothdeterministicandriskgenes

(adaptedfrom:(Guerreiro,Brasetal.2013,KarchandGoate2015)).

Numerousenvironmentalriskfactorsincludingdiet,exercise,education,andlifeactivitieshave

beenproposedinthelasttwodecadesastheresultofepidemiologicalinvestigations(Luchsinger,

Tangetal.2001,Scarmeas,Zarahnetal.2003,Scarmeas,Sternetal.2006,Luchsinger2008,

Scarmeas,Luchsingeretal.2009,Duncan,Nikelskietal.2017).Thoughthemechanisticimpact

of such factors remains unclear, there is growing consensus that they might impact one’s

4

‘cognitivereserve’whichisthoughttoprotectthebrainagainstAD-relatedpathology(Scarmeas,

Zarahnetal.2003,Scarmeas,Zarahnetal.2004,Stern2006).

1.1.3 PathologicalhallmarksADwasfirstdescribedover100yearsagobyGermanphysicianAloisAlzheimer.Whileexamining

the post-mortem brain of a patient with severe memory defects, he noticed abnormal

accumulationswhichhe termedplaquesand tangles (Figure1.2).Thesehallmarkshavesince

comprisedthegeneralhistopathologicalcriteriaforADdiagnosis.In1911,theamyloidnatureof

theseextracellularplaqueswasdescribedbytheGermanneuropathologistMaxBielschowsky,

butitwasn’tuntilthe1980sthatscientistsidentifiedbetaamyloid(Ab),thepeptideatthecore

ofamyloidplaques(Kirschner,Inouyeetal.1987).In1986,biochemicalstudiesuncoveredtau

(Mapt)asthepathologicalbuildingblockofneurofibrillarytangles(Grundke-Iqbal, Iqbaletal.

1986,Kosik,Joachimetal.1986).Debatelingersregardingwhichpost-translationallymodified

formscontributemosttotheseaberrantintracellularaccumulations(Grundke-Iqbal,Iqbaletal.

1986, Askanas, Engel et al. 1994, Novak 1994); however, it is largely accepted that

hyperphosyphorylationoftauisapathologiceventwhichpromotestangleformation(Alonso,

Zaidietal.2001,GongandIqbal2008,Wang,Xiaetal.2013,Hu,Zhangetal.2016).

5

Figure1.2ThepathologicalhallmarksofAD.

Silver staining of an AD brain reveals extracellular amyloid plaques (a) and intracellular

neurofibrillarytangles(b)(Serrano-Pozo,Froschetal.2011).

1.1.4 Diagnosticchallenges

AmajorhurdleinthetreatmentandresearchofADoverthepastseveraldecadeshasbeenthe

absenceofreliableclinicalbiomarkers.Commonpresentingsymptomsincludememorydeficits,

personalitychanges,anddisorientation(APA2017).Yet,notallpatientspresentwiththesame

constellation of symptoms and many of these symptoms are shared among other types of

transient or progressive neuropathies, rendering behavioral criteria alone insufficient. In the

early1990s,nearly90yearsafteritsrecognitionasadisease,acerebrospinalfluid(CSF)-based

biomarker profile (centering on Ab and tau levels) for AD provided amajor step forward in

antemortemdiagnosticsanddiseasemanagement.Tonote,itwasinitiallyhypothesizedthatCSF

fromADpatientswouldcontainhigherlevelsofAb,reflectingitsincreasedaccumulationinthe

brain;surprisingly,however,studiesrevealedthatADCSFcontainedstrikingly lessAb (Fagan,

Mintun et al. 2006),whichwas retrospectively attributed to its preferential sequestration in

plaques.

6

Recent technological advancements have paved the way for the development of diagnostic

biomarkersusingadditionalmodalities.Imagingbiomarkers,forexample,relyuponradiolabeled

compoundswhichcanbetargetedtowardsspecificbiologicalcomponentssuchasextracellular

amyloidandneurofibrillarytau.Acompoundnamed11C-PIBhasdemonstratedearlypromisein

thelabelingofamyloidplaques(Archer,Edisonetal.2006).Additionalplaque-targetingtracers

bearingthemorestableF18radioisotopehavealsobeendeveloped.Theseincludeflorbetapir,

florbetaben, flutemetamol, and the more recent fluselenamyl. Radioligands targeting

neurofibrillarytangles includeAV-1451(T807/flortaucipir),MK-6240,PI-2620,THK5317,PBB3,

andJNJ-067.Thegrowingarsenalofneuroimagingtoolstargetingthesetwohistopathological

hallmarksrendersimminenttheirinvivovisualization,whichcouldimprovediagnosticacuityand

histopathologicalmonitoring.

1.1.5 Theamyloidhypothesis

ThepredominanttheoryofADpathogenesis,termedtheamyloidhypothesis(oramyloidcascade

hypothesis),wasproposedin1991(HardyandAllsop1991)followingthediscoveryofdisease-

causingmutationsinAPPinearly-onsetAD(Chartier-Harlin,Crawfordetal.1991).Thistheory

positsthatincreasedaccumulationofextracellularAbpeptideaggregatesistheinitialpathogenic

eventwhichgivesrisetoalldisease-relatedneuropathology(HardyandAllsop1991).Despite

loomingspeculation,thishypothesishasremainedthedominanttheoryfornearlytwodecades,

motivatingextensivestudyoftheunderlyingbiochemistryandprovidingthefieldalong-awaited

moleculardrugtarget.

7

APP,theholoproteinofAb,isatype-Imembraneproteinhighlyexpressedinthebrain,whereit

isthoughttocontributetothestructuralandfunctionalintegrityofneuronalsynapses(Priller,

Baueretal.2006,Tyan,Shihetal.2012).Itcanbecleavedbyvariousenzymesalongtwogeneral

pathways, yielding fragmentswithdistinct functional properties. These twopathways are (1)

amyloidogenicand(2)nonamyloidogenic,namedfortheircapacitytogeneratetheAbpeptide

(Figure1.3).

Figure1.3Theprocessingpathwaysofamyloidprecursorprotein(APP).

TwopathwaysdefinetheprocessingfateofAPP.Thenon-amyloidogenicpathwayisinitiatedby

enzymaticcleavageofADAM9,ADAM10,orADAM17toyieldsAPPaandthe83-aminoacidC-

terminal fragment (C83ora-CTF).Amyloidogenic cleavagebeginswith cleavagebyBACE1 to

yield sAPPb and the 99-amino acid C-terminal fragment (C99 or b-CTF). This C99 fragment

8

undergoessubsequentcleavagebyag-secretasecomplex(composedofpresinilin1or2,nicastrin,

PEN2,andAPH-1)toyieldAbandtheAPPintracellulardomain(AICD)(LaFerla,Greenetal.2007).

TheamyloidogenicpathwaygivesrisetoAbviasequentialcleavageofAPPbyabeta-secretase

(BACE1)andag-secretase complex, yieldingapeptideofeither40aminoacids (Ab40)or42

aminoacids(Ab42)dependingonthesiteofgammacleavage.StudieshaveshownthatAb42is

more prone to aggregation (i.e. more pathogenic) than Ab40 and accordingly serves as the

centralmolecularculpritoftheamyloidhypothesis(Bitan,Kirkitadzeetal.2003).

Figure1.4Theamyloidcascadehypothesis.

This leading theory of ADpathogenesis asserts that the accumulation of Ab42 is the central,

initiatingeventwhichgivesrisetoallsubsequentpathology.Thoughvariousaggregateformsof

Ab42 may have differential pathologic effects, their accumulation is presumed causal of

9

downstreameventsincludingpairedhelicalformation(PHF)oftau,neuronalloss,andcognitive

decline(Karran,Merckenetal.2011).

Thisleadingtheory,informedbyagrowingbodyofresearch,assertsthattheaggregationofAb42

promotes a cascade of cytotoxic effects (Figure 1.4) which include tau dysregulation,

neurofibrillarytangleformation,andultimatelyneuronaldeath.Severaldetails,however,remain

unclear.Forexample,debatehasensuedregardingwhichaggregateformofAb42(monomeric,

oligomeric, or deposited plaque) contributemost to disease pathology (Walsh, Klyubin et al.

2002,Glabe 2006), and in general, themechanism throughwhich these events promote tau

dysregulation(Zheng,Bastianettoetal.2002,ManczakandReddy2013,Bloom2014).

The amyloid hypothesis is strengthened by several key findings. First, autosomal dominant,

disease-causingmutationsinAPPandgamma-secretasepresenilinsPSEN1andPSEN2havebeen

identified (Levy-Lahad,Wasco et al. 1995, Sherrington, Rogaev et al. 1995). The pathogenic

impactofthesemutationsisattributedtoincreasedAb42generationthrougheitheranincreased

productionofAb (bothAb40andAb42)altogetheroran increase in theratioofAb42/Ab40.

Second,Down’sSyndromepatientsareatsubstantiallyhigherriskforAD,withmorethanhalf

ultimately developing the disease. This has been attributed to the extra copy of APP (on

chromosome21), and therefore substantially elevated levels of itsmetabolic productswhich

include Ab (Teller, Russo et al. 1996). Third, genome wide association studies (GWAS) have

identifiedassociationsinseveralgeneswhichhavepurportedrolesinAbclearance(suchasAPOE

andclusterin)toAD(Zlokovic,Marteletal.1996,Tanzi,Moiretal.2004).Infact,APOEstatusis

10

currentlythestrongestgeneticriskfactorforsporadicAD,withtheAPOE4alleleaccountingfor

roughly half of all AD cases (Ashford 2004). Though the precise pathological contribution of

APOE4isdebated,studieshavedemonstrateditsdecreasedclearancecapacityforextracellular

AbascomparedwithAPOE2andAPOE3(Strittmatter,Weisgraberetal.1993,Bales,Verinaetal.

1999,Holtzman,Balesetal.2000,Shibata,Yamadaetal.2000,Holtzman2001,Ashford2004).

Yet, despite its rise as thepredominant theory ofADpathogenesis in the past twodecades,

several challenges to the amyloid hypothesis remain. First, there is an anatomicalmismatch

betweenextracellularplaquesandneurotoxicityinADbrains.Specifically,themedialtemporal

lobe, which undergoes the primary and most extreme neurotoxicity (Khan, Liu et al. 2014,

Altmann,Ngetal.2015)bears the leastplaqueburden in thecortex (BraakandBraak1991,

Altmann,Ngetal.2015).Second,theoveralldepositionofAbintheADbraindoesnotcorrelate

with cognitive decline; in fact, extensive amyloid plaques have been documented in non-

demented, cognitively ‘normal’ individuals (Rentz, Locascio et al. 2010). Third, other

histopathological events such as endocytic abnormalities appear to precede local amyloid

depositioninADbrains(Cataldo,Peterhoffetal.2000),raisingquestionastothepositioningof

amyloidaccumulationalongthepathologicaltimelineofAD.Andfourth,todate,agrowinglist

oftherapiestargetingextracellularamyloidhavelargelyfailedtoshowefficacyintreatingthe

disease. One must consider that the failure of Ab-targeting therapies could derive from

neurodegenerationoccurringpriortointervention.However,takentogether,theseobservations

pose challenge to the predominant theory of AD pathogenesis and motivate the quest for

additionaldrugtargets.

11

1.1.6 TheendosomeasanemergingtherapeutictargetThoughnot considereda canonicalbiomarkerofAD,endosomaldysfunctionhas increasingly

beenrecognizedasasalientfeatureofitscytopathology.Cataldo,etal.describedanabnormal

enlargement (2.5-fold) of neuronal endosomes as an early event in AD pathology (Cataldo,

Barnettetal.1997)andlaterreportedthatthiscellulareventprecedeslocalizedAbdeposition

inbothsporadicADandDownSyndromepatients(Cataldo,Peterhoffetal.2000).Othergroups

havealsodescribedendosomalenlargementasafeatureofIPSC-derivedneuronsfrompatients

with both sporadic (Israel, Yuan et al. 2012) and autosomal-dominant forms of AD (Raja,

Mungenastetal.2016).

Furthermore, genetic studies have linked a growing number of genes involved in endosomal

traffickingtoAD.TheseincludeSORL1,BIN1,PICALM,ABCA7,CD2AP,SNX1,SNX3,RAB7A,and

KIAA1033(Rogaeva,Mengetal.2007,Vardarajan,Bruesegemetal.2012,Guerreiro,Brasetal.

2013,Lambert,Ibrahim-Verbaasetal.2013,KarchandGoate2015,Steinberg,Stefanssonetal.

2015).As such, endosomal traffickinghas risenasoneof themajor categoriesofAD-related

genes (Small, Simoes-Spassov et al. 2017) (Figure 1.5). And, asmentioned previously, causal

mutationsintheendosomaltraffickingreceptorSORL1havebeenidentified(Pottier,Hannequin

etal.2012,Vardarajan,Zhangetal.2015,Verheijen,VandenBosscheetal.2016),underscoring

apathogenicroleforendosomaldysfunctioninAD.

12

Figure1.5EndosomaltraffickinginAD.

(a)Geneticstudieshaveidentifiedvariantsinagrowingnumberofgeneswhichconferincreased

risk of AD. These genes broadly coalesce into four functional categories including endosomal

trafficking. As studies have shown defects in the other three gene categories can lead to

endosomaltraffickingdefects,thiscategoryrepresentsacommonfinalpathwayfordownstream

neurotoxicevents.(b)EndosomaltraffickinggeneslinkedtoADperformdistinctfunctionswhich

canaffectsortingdynamicsattheendosomalhub.Endosomaltrafficjams,nowthoughttoplay

a role in AD pathogenesis, result from defects to the degradation pathway (1), the recycling

pathway(2),and/ortheretrogradepathway(3)(Small,Simoes-Spassovetal.2017).

13

Finally, expression profiling studies of post-mortem brain tissue have identified a significant

reductionofendosomaltraffickingproteins(VPS35,VPS26a,andBeclin-1)andlipids(PI3P)inthe

areasof thebrainmostaffected inAD (Small,Kentetal.2005,Pickford,Masliahetal.2008,

Morel,Chamounetal.2013).Thoughexpressionprofilingofpost-mortembrainsisoftenlimited

tolatestageAD,Beclin-1reductionwasdescribedinapost-mortemstudyofpatientsintheinitial

symptomaticstageofmildcognitiveimpairment(MCI),positioningitsdeclineasanearlyevent

inADpathology(Pickford,Masliahetal.2008).

ThegeneticandbiochemicalstudieslinkingendosomaltraffickingdeficitstoADhavemotivated

mechanisticexplorationoftheresultingmolecularpathology.Asaresultofthisresearch,there

is growing consensus that endosomal dysfunction can promote Ab accumulation through a

varietyofmechanisms(Wen,Tangetal.2011,Morel,Chamounetal.2013,Nixon2017,Small,

Simoes-Spassovetal.2017);theseincludetraffickingdefectswhichincreasetheretentiontime

ofAPPand/or itsamyloidogeniccleavageenzymesintheendosome,therebypromotingtheir

interaction(Wen,Tangetal.2011,Bhalla,Vetanovetzetal.2012).Interestingly,thereversehas

also been demonstrated: that amyloidogenic APP processing can promote endosomal

dysfunctionduetocytotoxiceffectsofamyloidogenicAPPproducts(Jiang,Mullaneyetal.2010,

Xu, Weissmiller et al. 2016, Small, Simoes-Spassov et al. 2017). Accordingly, endosomal

dysfunctionhasbeenproposedasacommoncellularpathwaycapableofmediatingsubsequent

AD-relatedpathology(Small,Simoes-Spassovetal.2017)(Figure1.5)andanoveldrugtargetfor

discovery(Berman,Ringeetal.2015,Small,Simoes-Spassovetal.2017).

14

1.2 Theretromerassembly

1.2.1 Retromerasamasterconductorofendosomaltrafficking

Retromerhasbeentermedamasterconductorofendosomaltrafficking,asitcoordinatesthe

transportofproteincargooutoftheendosomealongtwopathways:theretrogradepathway

towardthetrans-golginetwork(TGN)ortherecyclingpathwaytowardtheplasmamembrane

(Figure1.6a).Retromerwasfirstdescribedinyeastininthelate1990s(Seaman,Marcussonet

al. 1997, Seaman, McCaffery et al. 1998); its mammalian orthologs were identified shortly

thereafter (Edgar and Polak 2000, Haft, de la Luz Sierra et al. 2000) and have been studied

extensivelyintheyearssince.

Earlystudiesinyeastrevealedacriticalroleforretromerinendolysosomalhealth.Specifically,

Seaman et al. demonstrated that retromer traffics receptor VPS10p along the retrograde

pathway,ensuringcontinuousdeliveryofitsligandcarboxypeptidaseY(CPY)fromtheTGNto

theprevacuolarendosome (Seaman,Marcussonet al. 1997, Seaman,McCafferyet al. 1998).

They further showed that mutations in retromer caused an increased vacuolar retention of

VPS10pandanabnormalsecretionofvacuolarproteaseCPYintotheextracellularspace.Follow-

upstudiesconfirmedasimilarroleforretromerinthemammaliansystem:lysosomaldeliveryof

proteasecathepsinDviaretrogradetraffickingofitsreceptor,thecation-independentmannose-

6-phosphatereceptor(CIM6PR)(Arighi,Hartnelletal.2004).Numerousstudiesintheyearssince

have unveiled a multitude of other retromer cargo including mammalian VPS10 domain-

15

containingreceptors(e.g.SORL1)andglutamatereceptors(e.g.GluR1)(SmallandGandy2006,

Zhao,Nothwehretal.2007,Lane,Rainesetal.2010).

Theterm‘retromer’refersnotsimplytooneprotein,butinsteadamulti-modularassemblyof

proteins which must organize biochemically to traffic endosomal cargo to the appropriate

destination(SmallandPetsko2015).Thesemodulesofretromerassemblyhavebeengrouped

into five categories (Figure 1.6b) which coalesce around their functional contribution: cargo

recognition,membranerecruiting,tubulation,actin-remodeling,andthecargotobetrafficked

toitsdestination(SmallandPetsko2015).

Figure1.6Theretromerassembly.

(a)Retromertrafficscargointwoofthethreeroutesoutoftheendosome:therecyclingpathway

and the retrograde pathway. (b) The five modules of retromer assembly include the cargo

16

recognitioncore (which includesVPS35,VPS29,andeitherVPS26aorVPS26b), themembrane

recruitingmodule(whichincludesbothproteinsandlipidssuchasPI3P),thetubulationmodule,

actin-remodelingcomponents,andthecargoandreceptorstobetraffickedoutoftheendosome

(adaptedfrom(SmallandPetsko2015)).

Thoughindividualcomponentsofthisassemblymayvaryinacontext-dependentmanner,the

cargorecognitioncore,whichconsistsof3vacuolarproteinsortingproteins(VPS35,VPS29,and

either VPS26a or VPS26b), is an essential and unifying component of the retromer assembly

(Figure1.6b).Infact,expressionofthesecoreproteinsisoftentightlyregulatedwithinthecell,

andadeficiency inanyoneof theseproteins results in concomitant reductionof its trimeric

bindingpartners(unpublisheddata).RecentstudiessuggestthatcoreproteinparalogsVPS26a

andVPS26bformdistincttrimerswithVPS35andVPS29andthattheseassembliesmayperform

differentfunctionsinthebrain(Kerr,Bennettsetal.2005,Kim,Leeetal.2010,Bugarcic,Vetter

etal.2015).

This core trimer is recruited to theendosomalmembrane via componentsof themembrane

recruitingmodule,whichcanincludeproteinssuchassortingnexin-3(SNX3)(Harterink,Portet

al.2011)andRAB7A(Rojas,vanVlijmenetal.2008,Seaman,Harbouretal.2009,Harrison,Hung

et al. 2014). The endosomal-enriched lipid phosphatidylinositol-3-phosphate (PI3P) also

contributestomembranerecruitmentoftheretromerassemblythroughbindingofSNXsviatheir

phoxhomologydomain(Cozier,Carltonetal.2002)andisthereforeincludedinthisfunctional

category.

17

Essential also for this trafficking process is the generation of a nascent tubule toward the

directionofcargotransport.Thisfunction isperformedbythemembranetubulationmodule,

comprisedofsortingnexinsSNX1,SNX2,SNX5,andSNX6(Rojas,Kametakaetal.2007,Wassmer,

Attaretal.2007).Heterodimersofthesesortingnexinsdirectlybindthecargorecognitioncore

while stabilizing membrane curvature via carboxy- terminal BIN–amphiphysin–RVS (BAR)

domains(AttarandCullen2010,MimandUnger2012).

Anactin-remodelingmodulebindstotheretromerassemblyviaVPS35whereitgivessupportto

thebuddingtubuleviathegenerationofactinpatches(Harbour,Breusegemetal.2010).The

proteins involved inthisprocess includemembersof theWASHcomplex:WASprotein family

homologue1(WASH1),FAM21,strumpellin,coiled-coildomain-containingprotein53(CCDC53)

andKIAA1033(GomezandBilladeau2009,DeriveryandGautreau2010,Harbour,Breusegemet

al.2010).

The fifth module of retromer assembly consists of the cargo and/or cargo receptors to be

transportedoutoftheendosome.Whilevariouscargoofretromercontinuetobeunearthed,

severalbearingneurologicalrelevancehavebeenidentifiedtodate.TheseincludeVPS10-domain

containingreceptors(SORL1,SORCS1,SORCS2,SORCS3,andsortilin),severalofwhichhavebeen

recognizedfortheirroleinthetransportofAPPoutoftheendosome(Small,Kentetal.2005,

Rogaeva,Mengetal.2007,Bhalla,Vetanovetzetal.2012,Fjorback,Seamanetal.2012,Lane,St

George-Hyslop et al. 2012). Asmentioned above, CIM6PR,whichdelivers cathepsinD to the

18

endolysosomalsystem,hasalsobeenverifiedasaretromercargo(Arighi,Hartnelletal.2004,

Seaman2004). Finally, severalneuronal signaling (GluR1 (Temkin,Morishitaetal.2017))and

microgliaphagocytic (CD36andTREM2 (Lucin,O'Brienetal.2013)) receptorshavealsobeen

showntorelyonretromerfortransportalongtherecyclingpathway(Lucin,O'Brienetal.2013,

Choy,Parketal.2014).

1.2.2 RetromerinAD

Retromerwas first linked to AD by our lab in 2005 through amicroarray experimentwhich

revealed that the entorhinal cortex of AD brains are deficient in retromer core components

VPS35andVPS26(Small,Kentetal.2005).ThoughthereasonfordecliningretromerlevelsinAD

wasunclearatthetime,recentworksuggeststhatbothgeneticsandenvironmentcanplayarole

(Rogaeva,Mengetal.2007,Morabito,Bermanetal.2014).

Follow-upstudiesalsorevealedthatsilencingVPS35invitroincreasesAbproduction,suggesting

that this defect might be upstream of AD pathology (Small, Kent et al. 2005). A cellular

mechanismwas lateruncovereduponthe identificationofAPPreceptorSORL1asaretromer

cargo,highlightingtheamyloidogeniceffectsofextendedendosomalretentionofAPP(i.e.that

traffickingmatters)(SmallandGandy2006,Muhammad,Floresetal.2008,SmallandDuff2008,

Lane,Rainesetal.2010,Lane,StGeorge-Hyslopetal.2012).Follow-upstudieshaveconfirmed

that retromer deficiency (via knockdown of VPS35 or VPS26a) can lead to increased

amyloidogenic processing through trafficking defects in APP (Muhammad, Flores et al. 2008,

Bhalla,Vetanovetzetal.2012,Choy,Chengetal.2012).Interestingly,retromerhasalsobeen

19

showntotrafficBACE1outoftheendosome(Wen,Tangetal.2011)(whereitisthoughttoimpart

themajority of its amyloidogenic effects on APP (Small and Gandy 2006)). This suggests an

additional mechanism through which retromer dysfunction enhances Ab accumulation,

increasingretentionoftheinitiatingamyloidogenicenzymeaswellasitssubstrateAPPinthe

neuronalendosome.

GeneticsstudieshavefurthersupportedapathologicalroleofretromerinAD.Todate,variants

havebeenidentifiedingeneticcomponentsofall5retromermoduleswhichconferanincreased

riskforAD(Rogaeva,Mengetal.2007,Vardarajan,Bruesegemetal.2012,Lambert, Ibrahim-

Verbaasetal.2013).Asmentionedpreviously,causalmutationsinretromerreceptorSORL1has

alsobeendiscovered(Pottier,Hannequinetal.2012,Vardarajan,Zhangetal.2015),positioning

retromerdysfunctionupstreamoftheresultingpathologyinatleastasubsetofcases.

1.2.3 RetromerdeficiencyasamodelofAD-relatedendosomaldysfunction

RetromerdeficiencyinvitrorecapitulatesseveralaspectsofAD-relatedendosomaldysfunction.

First, knockdown of retromer core proteins promotes Ab accumulation through increased

retentionofAPPandBACE1intheendosome(Wen,Tangetal.2011,Bhalla,Vetanovetzetal.

2012). Second, it results in an enlargement of endosomes (Offe, Dodson et al. 2006, Bhalla,

Vetanovetzetal.2012),purportedlythroughanaccumulationofcargoatthelimitingmembrane.

Third, retromer deficiency disrupts delivery of protease cathepsin D to the endolysosomal

system,resultinginitsincreasedextracellularsecretion(andincreasedcathepsinDsecretioninto

theCSFhasbeenreported inADbrains) (Schwagerl,Mohanetal.1995).Fourth, itcausesan

20

increaseinfreecholesterol(alsoreportedinAD(Cutler,Kellyetal.2004,Xiong,Callaghanetal.

2008)),anditsredistributiontolateendocyticcompartments(Marquer,Tianetal.2016).Finally,

recentworkinhumaniPSC-derivedneuronsdemonstratesthatretromerdeficiencypromotes

hyperphosphorylationoftau,andthatthiseffectisindependentofitseffectsonAb(Young,Fong

etal.2018).

Mousemodelsofretromerdeficiency,generatedbyselectivedeletionofthecargorecognition

core,alsoexhibitenhancedaccumulationofAbinthebrain(Muhammad,Floresetal.2008,Wen,

Tangetal.2011).Furthermore,theydemonstratedefectsinlong-termpotentiation(whichmay

be reversed using gamma-secretase inhibitors (Muhammad, Flores et al. 2008)) and

hippocampal-dependentmemorytasks(Muhammad,Floresetal.2008,Wen,Tangetal.2011).

1.3 Thesearchforbiomarkersofendosomaldysfunction

1.3.1 Rationale

Endosomal dysfunction is clearly linked toADpathology, and represents a novel therapeutic

targetfordiscovery(Berman,Ringeetal.2015,Small,Simoes-Spassovetal.2017).Criticaltothis

pursuit,however, istheidentificationofabiomarkerofendosomaldysfunction,servingthree

primarypurposes:1)addressingtheepidemiologyofendosomaldysfunctioninAD;2)tracking

thediseasecourse;and3)properlystratifyingandmonitoringpatientsinupcomingclinicaltrials.

Theprojectdescribedhereinwasdesignedtoaddressthisunmetneedthroughidentificationof

aclinicallyrelevantbiomarkerofAD-relatedendosomaldysfunction.

21

1.3.2 Hypothesis

ThegeneralhypothesismotivatingthisinvestigationisthatAD-relatedendosomaldysfunctionin

themurinebrainwillimpartmolecularchangesintheextracellularmilieuwhichcanbedetected

viaclinicallyusefulbiofluids.Prior invitrostudieshavedemonstratedthatendosomaldefects

canpromotebothproteinandlipidalterationsinthecellularsecretome(Rojas,vanVlijmenetal.

2008,Yuyama,Yamamotoetal.2008,Sullivan,Jayetal.2011).Toourknowledge,thisstudyis

thefirsttofullytranslatesuchlogicintoadiscovery-driveninvestigationofAD-relatedendosomal

dysfunctioninthemurinebrain.

1.3.3 Strategy

Ourstudydesignbroadlyinvolvesthegenerationofmousemodelsofendosomaldysfunction,

thecollectionandanalysisofbiologicalreservoirsmolecularbiomarkersusingunbiasedscreens,

andultimatelytheirtranslationtohumansubjects(Figure1.7).Thisstrategyisoutlinedinmore

detailinthesectionsthatfollow.

Figure1.7ThesearchforbiomarkersofAD-relatedendosomaldysfunction.

22

Thisproject involves the collectionofCSFandbrainexosomes from retromer-deficientmouse

modelsforinvestigationofproteomicandlipidomicchangeswhichcanbetranslatedtohuman

subjects. Utility of such a biomarker includes uncovering the epidemiology of endosomal

dysfunctioninAD,aswellasproperpatientstratificationandmonitoringforupcomingclinical

trialsofendosome-targetingagents.

1.3.3.1 Techniques

WehaveoptedtoinvestigatemolecularchangesthatariseinboththeCSFandinpurifiedbrain

exosomes fromourmousemodelsof endosomaldysfunction.CSF is an intuitive reservoir as

retromerdysfunctionhasbeenshowntoimpactthecellularsecretome(e.g.enhancedsecretion

ofcathepsinDandAb,proteinswhicharereadilydetected inCSF) (Muhammad,Floresetal.

2008,Rojas,vanVlijmenetal.2008,Bhalla,Vetanovetzetal.2012).Itisalsonowatechnically

feasiblehaven fordiscoveryasprotocols formurineCSFcollectionhavebeendevelopedand

successfullyreportedinagrowingnumberofstudies(Barten,Cadelinaetal.2011,Schelle,Hasler

et al. 2017). Biomarkers identified in this fluid could therefore directly translate to human

subjectsasCSFcollectioninhumansisawell-establishedclinicalprocedure.

Brain-derivedexosomesrepresentthesecondreservoirfordiscoveryinthisproject.Asexosomes

derivefromthelimitingendosomalmembrane(whichisalteredinretromerdysfunction)(Bhalla,

Vetanovetz et al. 2012), it follows that these changes might be reflected in the secreted

exosomes.Infact,invitrostudieshaveconfirmedelevationofC-terminalfragmentsofretromer

cargoAPPinexosomesfollowingknockdownofcoreproteinVPS35(Sullivan,Jayetal.2011).The

23

recent development of two critical techniques position brain-derived exosomes as a feasible

havenforbiomarkerdiscovery.Thefirst isaprotocol for isolatingexosomesfromthemurine

brain(Perez-Gonzalez,Gauthieretal.2012).Thesecondisaprotocolforisolatingbrain-derived

exosomesfromhumanserum(Goetzl,Boxeretal.2015,Goetzl,Mustapicetal.2016).Together

thesemakepossibletheindirecttranslationfromfindingsinmurinebrainexosomestohuman

brain-derivedserumexosomes,whichcanbeobtainedclinicallythroughvenipuncture.

Wehave chosen to investigate both protein and lipid changes as both can be altered in the

cellularsecretomeasaresultofendosomaldysfunction(Rojas,vanVlijmenetal.2008,Yuyama,

Yamamotoetal.2008,Strauss,Goebeletal.2010,Sullivan,Jayetal.2011,Bhalla,Vetanovetzet

al.2012).Thesemolecularapproacheswillbedescribedfurtherintheensuingchapters.

1.3.3.2 MouseModels

We have amassed an assortment of retromer-deficientmousemodels, distinct in either the

deletedcargorecognitioncoreproteinorintheregionofgeneticdefect(Figure1.8).TheCamk2a

knockoutmodels(VPS35andVPS26a),restrictretromerdeficiencytoforebrainneurons.Andthe

useofVPS26a inparticular restricts themoleculardefect to theVPS26a-containingassembly

(whileVPS35depletionaffectsbothVPS26a-andVPS26b-containingassemblies).Knockoutof

VPS26b—thebrain-enrichedVPS26paralog(Kerr,Bennettsetal.2005)—providesanothermodel

ofregionalretromerdeficiency(inthiscase,ofonlytheVPS26b-containingassembly).Finally,

useoftheVPS26a-haploinsufficient(knockoutisembryoniclethal)modelallowsexplorationof

retromerdeficiencyinallbraincells,asstudieshavesuggestedthatdeficiencyoftheVPS26a-

24

containingretromerassemblyinADmaynotbeentirelylimitedtoneurons(Lucin,O'Brienetal.

2013).

Figure1.8Mousemodelsofretromer-dependentendosomaldysfunction.

(a)Retromer-deficientmodelshavebeengeneratedbydeletionofcargorecognitioncoreproteins

VPS35,VPS26a,orVPS26b.(b)Thesemodelsdiffernotonlyinthechoiceofcoreprotein,butalso

thelevelandselectivityofdeletion.

Finally,ourdesignincludesthetranslationofthemostcompellingCSForexosomalfindingsto

humansubjects,eitherdirectly inthecaseofCSForviaextractionofbrain-derivedexosomes

fromhumanserum(Goetzl,Boxeretal.2015).

We acknowledge that the analysis of two classes of molecules (lipids and proteins) in two

biologicalreservoirs(CSFandexosomes)forthefourretromer-deficientmodelslistedaboveis

anambitiousundertaking,requiringsubstantialtime,materials,andanimals.Assuch,wehave

chosenaself-refiningapproach,limitingourefforttothosestudiesdeemedthemostfruitfuland

eliminatingthecontinueduseof techniquesoranimalmodelswhichcarry limitedpromiseof

25

biomarkerdiscovery.Thus,notalltechniqueswereperformedforallrodentmodels,andnotall

findingswerefullytranslatedtohumansubjects.

2 Chapter2:CSFLipidomics

2.1 Introduction

2.1.1 Lipidsandtheendosome

Lipids comprise a diverse class of molecules which are critical for many aspects of cellular

function including membrane formation, energy metabolism, signaling, and intracellular

trafficking.Furthermore,lipidsformtheenclosingstructurefornotonlycellularentitiesbutan

entirehostofmicrovesicles(e.g.ectosomes,exosomes)theyreleaseintotheextracellularspace.

Infact,studieshavedemonstratedthatgeneticdefectswhichimpedeendolysosomaldynamics

canaffectnotonlyinternallipidmetabolism(Chen,Gordonetal.2005),butalsothepopulation

of secreted lipids (Strauss, Goebel et al. 2010), thus providing an extracellular signal of

intracellularendosomaldysregulation.Itfollows,therefore,thatretromerdysfunctionmaylead

toalterationsinintracellularlipidlocalization,andultimatelyextracellularsecretion.

2.1.2 LipiddysregulationinAD

LipiddysregulationhasbeenobservedinvariouspathologiesincludingAD(Foley2010,DiPaolo

and Kim 2011,Morel, Chamoun et al. 2013). These observations include altered cholesterol

metabolismanddeficienciesofendosomaltraffickingsubstituentssuchasPI3P(Foley2010,Di

PaoloandKim2011,Morel,Chamounetal.2013),anessentialcomponentforretromerassembly

26

(Rojas,vanVlijmenetal.2008).Assuchdysregulationcanresultfromendosomaldysfunction

(Marquer,Tianetal.2016), it is reasonable topostulate that thesetwoobservationsmaybe

linkedintheADbrain.

2.1.3 Massspectrometryoflipids

Theadventofmassspectrometryinrecentdecadeshaspavedthewayforlipidomicinvestigation

ofcells,tissues,andbiofluids,andhasfurtheredourunderstandingoftheirbiologicalrelevance.

LC-MS allows both qualitative and quantitative analyses and has shown great promise in

biomarker discovery of various biological fluids (Capodivento, Visigalli et al. 2017, Zarrouk,

Debbabietal.2017).

2.1.4 CSFlipidomicsstrategy

CSF collection inmice is not trivial. Roughly 5uL of fluid can be extracted from eachmouse

throughthecisternamagnaviaapost-mortemtechnique,thoughtheamountcanvarybasedon

amultitudeoffactorssuchasage,gender,anddiet.Specialcareisrequiredtominimizeblood

contamination as an intricate network of blood vessels overlay this tiny pocket of fluid, and

substantialpracticeiswarrantedtoattaintherequisiteprecisionandconsistencyformolecular

profiling.Accordingly,thefirststepofthisprojectinvolvedmasteryofhighqualitymurineCSF

collection,whichisreportedintheresultsthatfollow.

Uponcommencementofthisproject,onlyahandfulofmurineCSFstudieshadbeenpublished,

andnoneincludedlipidomics.Therefore,consideringtherelativelylowconcentrationoflipidsin

27

humanCSF(roughly10-13ug/mL)(Seyer,Boudahetal.2016),weoptedforapoolingapproach,

intendedtomaximizetheamountoffluidfordiscoverywhilestilloperatingwithinthelimitsof

ourmousehusbandrycapacity.CSFcollectionfromthisretromermodelyieldedroughly5-10uL

permouse(versusseveralmillilitersfromhumans).Here,wechosetopoolsamplestoafinal

volume of 30uL per biological replicate in order to boost the signal/noise ratio for the less

abundant lipid species and ultimately expand the array of detectable lipids available for

biomarkerdiscovery.

2.1.5 Choiceoftheretromer-deficientmodel

We began our studies with the VPS26a-haploinsufficient mice for several reasons. First,

haploinsufficiency is likelymorereflectiveof theextentof retromerdeficiency inADpatients

thanacompleteknockoutofretromercoreproteins.Second,asretromerdeficiencyinADmay

notbelimitedtoneurons(Lucin,O'Brienetal.2013),wechosetobeginwithamodelinwhich

thegeneticdeficitissharedamongallcelltypesinthebrain.

2.2 Results

2.2.1 CollectionofhighqualityCSF

Murine CSF collection is a highly-skilled procedure which requires much practice to yield

consistentresults.InordertoperformlipidomicinvestigationsofmurineCSF,wefirstneededto

ensure that we could collect high quality samples from our rodent models. As blood

contaminationcouldhavetremendousimpactontheresultsofouranalyses(duetothedistinct

28

molecular profiles of these twobiofluids),we implemented a sensitive hemoglobin ELISA for

qualityassessmentofindividualsamples.

CSFwascollectedfrom40agoutimiceusingapost-mortemprocedure(Figure2.1)andassessed

forquality.PreviouslypublishedbloodcontaminationthresholdsformurineCSFstudiesinclude

0.003%(Barten,Cadelinaetal.2011)androughly0.01%(Dislich,Wohlrabetal.2015)(reported

threshold for visible contamination). Using these two criteria we are able to isolate

contamination-free CSF in these mice in 87% of samples (at a threshold of 0.003% blood

contamination)or95%(atathresholdof0.01%bloodcontamination).Samplespassingthemost

stringentofthresholdswasusedfordownstreamanalyses.

Figure2.1CollectionofhighqualityCSF.

(a)MurineCSFwascollectedthroughthecisternamagnainapost-mortemprocedure.(b)Forty

sampleswereobtainedfromagoutimiceandassessedforqualityusingtwobloodcontamination

thresholdswhichhavebeenpreviously reported:0.003%and0.01%(Levy-Lahad,Wascoetal.

1995,Scarmeas,Sternetal.2006,Barten,Cadelinaetal.2011,Dislich,Wohlrabetal.2015).

29

2.2.2 ThemurineCSFlipidomecloselyresemblesthatofhumans

Utilityofthisstrategyforpurposesofbiomarkerdiscoveryhingedgreatlyuponbothourability

todetectlipidsinasmallvolumeofCSFandonsubstantialsimilarityinthecompositionofthe

mouseandhumanCSFlipidome.Asmurinelipidomicstudieshadnotyetbeenreportedatthe

timeofthisstudy,wefeltitimperativetofirstcharacterizethemurineCSFlipidomeandcompare

itscompositiontothatofhumanCSF.

Tothisend,weperformedapilotofLC-MSonlipidsextractedfrombiologicalreplicatesof30uL

ofpooledwildtypeagoutiCSF.First, it isworthnoting thatwewereable todetect signal for

nearlyalllipidclassesthatarepresentinhumanCSFandthattheiroverallcontributiontothe

CSF lipidome is remarkably similar to that of humans (Figure 2.2a; results from independent

runs).Thelimitedvolume(30uLformiceversus300uLforhumans),however, largelyrestricts

discovery to the most highly abundant lipid species (cholesteryl esters, free cholesterol,

phosphatidylcholine, sphingomyelins, diacylglycerol, triacylglycerol, and

lysophosphatidylcholine). And often, within each lipid class, not all individual lipid species

detectableinhumanCSFaredetectableinsuchasmallvolumeofwildtypemouseCSF(Figure

2.2b).

30

Figure2.2ComparisonofthemouseandhumanCSFlipidome.

(a)HighqualityCSFfromagoutiwildtypemiceunderwentlipidextractionandwasrunviaLC-MS

in parallel (using six pooled biological replicates of 30uL each); internal standards provided

31

absolutequantitation.Separately,human(non-dementedcontrol)CSFsampleswererunviaLC-

MSinparallelusing300uLpersample.Lipidclassesarereportedasapercentageoftotaldetected

lipid species (mol %); FC=free cholesterol; CE=cholesteryl ester; AC= acyl carnitine ;

MG=monoacylglycerol; DG=diacylglycerol; TG=triacylglycerol; Cer=ceramide; dhCer=

dihidroceramide;SM=sphingomyelin;dhSM=dihidrosphingomyelin;GalCer=galactosylceramide;

Sulf=sulfatide;PA= phosphatidicacid;PC=phosphatidylcholine;PCEe=phosphatidylcholineether;

PS=phosphatidylserine; PI=phosphatidylinositol; PG=phosphatidylglycerol;

BMP=bismonoacylglycerophosphate; AcylPG=acyl-phosphatidylglycerol;

LPC=lysophosphatidylcholine; LPCe=lysophosphatidylcholine ether; NAPE=N-acyl

phosphatidylethanolamine;NSer=N-acylserine.(b)Lipidclassesarefurtherbrokendowninto

theindividualmoleculeswhichcomprisethatclass.Oftheeighteencholesterylestersidentified

inhumanCSF, theeightmosthighlyabundantwerealsodetected in30uLofmurineCSF (red

boxes),suggestingahighlevelofsimilarityinthemurineandhumanCSFlipidome.

2.2.3 LipidomicscreenrevealssphingomyelinchangesintheCSFofVPS26a-haploinsufficient

mice

We therefore applied this pooling approach to our investigation of CSF lipid changes in the

VPS26a+/-mice(Figure2.3).HighqualityCSFfrom3-month-oldVPS26a+/-malesandlittermate

controlswere pooled into six biological replicates. Following lipid extraction and addition of

internalstandards,LC-MSwasperformedforalltwelvereplicatesinparallel.

32

Figure2.3VPS26a+/-CSFlipidomicsstudyoverview.

CSFwascollectedfromVPS26a+/-miceandcontrollittermates.Contamination-freesampleswere

pooledintobiologicalreplicatesandrunviaLC-MSfollowinglipidextractionandtheadditionof

internalstandardsforabsolutequantitation.

AglobalcomparisonofthegenerallipidclassesrevealednosignificantchangesintheCSFfrom

3-month-oldVPS26a+/-miceascomparedwiththeirlittermatecontrols(Figure2.4a).However,

acloserlookattheindividuallipidspeciessuggestedalteredsphingomyelinsecretionintothe

CSFbytheVPS26a-haploinsufficientbrain(Figure2.4b).Inparticular,foursphingomyelinspecies

(d18:1/18:1,d18:1/22:1,d18:1/24:1,andd18:1/26:0)weredecreasedinCSFfromtheVPS26a

knockdownmice.Interestingly,SMd18:1/18:1wasabletodistinguishthetwogenotypeswitha

92%diagnosticaccuracy(p<0.01)inaregressionmodel.

33

Figure2.4LipidomicscreenrevealsalteredsphingomyelinprofileinVPS26a+/-CSF.

(a)VPS26ahaploinsufficiencyimpartsnosignificantchangestotheclassesofCSFlipids.Values

arenormalizedtototallipidcontentandthentothewildtypeandrepresentthemean+/-SEM

(n=6); FC=free cholesterol; CE=cholesteryl ester; AC= acyl carnitine ; MG=monoacylglycerol;

DG=diacylglycerol; TG=triacylglycerol; Cer=ceramide; dhCer= dihidroceramide;

SM=sphingomyelin; dhSM=dihidrosphingomyelin; GalCer=galactosylceramide; Sulf=sulfatide;

PA= phosphatidic acid; PC=phosphatidylcholine; PCEe=phosphatidylcholine ether;

PS=phosphatidylserine; PI=phosphatidylinositol; PG=phosphatidylglycerol;

BMP=bismonoacylglycerophosphate; AcylPG=acyl-phosphatidylglycerol;

LPC=lysophosphatidylcholine; LPCe=lysophosphatidylcholine ether; NAPE=N-acyl

34

phosphatidylethanolamine; NSer=N-acyl serine. (b)Abreakdownof individual sphingomyelin

speciesrevealsdecreasesinSMd18:1/18:1,d18:1/22:1,d18:1/24:1,andd18:1/26:0.Valuesare

normalizedtototallipidcontentandthentothewildtypeandrepresentthemean+/-SEM(n=6;

*P-values<0.5)

WesoughttovalidatethesefindingsfromourinitialscreeninanewcohortofmiceusingCSF

fromindividualsamples.Here,weselectedamoretargetedapproach,assessingonlypositive-

modelipidswhichincludesthesphingomyelinclass.Thisallowedustoconcentratesignalinthe

lipid category of interest instead of splitting into three modes. We were hopeful that this

approach would compensate for the smaller volume available for this follow-up validation.

Despite our efforts, however, we were unable to detect two of the sphingomyelin species

(d18:1/26:0 and d18:1/26:1) from the first cohort of pooled CSF. Among the remaining ten,

d18:1/18:1 was once again markedly reduced in the retromer-deficient CSF, confirming its

decreaseintheVPS26a-deficientbrain,whilemostremainingsphingomyelinspeciesshoweda

trendtowarddecrease(Figure2.5).

35

Figure2.5Follow-upstudyinnewcohortconfirmsdecreaseofSMd18:1/18:1inVPS26a+/-CSF.

Afollow-upanalysisofCSFfromindividual3-month-oldVPS26a+/-miceandlittermatecontrols

was conducted. (a)No changes in the positivemode lipid classeswere observed. Values are

normalizedtototallipidcontentandthentothewildtypeandrepresentthemean+/-SEM(n=6);

Cer=ceramide; SM=sphingomyelin; dhSM=dihidrosphingomyelin; GalCer=galactosylceramide;

PC=phosphatidylcholine; PCEe=phosphatidylcholine ether; LPC=lysophosphatidylcholine;

LPCe=lysophosphatidylcholineether.(b)Abreakdownofindividualsphingomyelinspeciesreveals

decreasesinSMd18:1/18:1,d18:1/22:1,d18:1/24:1,andd18:1/26:0.Valuesarenormalizedto

totallipidcontentandthentothewildtypeandrepresentthemean+/-SEM(n=6;*P-values<0.5)

36

(b) Among the ten sphingomyelins detected in these samples, SM d18:1/18:1 was again

significantlyreducedinVPS26a+/-CSF.Valuesarenormalizedtothewildtypeandrepresentthe

mean+/-SEM(n=8;*P-value<0.05).

2.2.4 ReductionsinVPS26a+/-CSFsphingomyelinsaremorepronouncedwithageandextend

tootherlipidgroups

CuriousastowhateffectagemighthaveontheCSFsphingomyelinprofile inthesemice,we

repeatedthisexperimentusing individualsamples in6-month-oldmice.Again,sphingomyelin

d18:1/18:1 was reduced in the CSF, as well as d18:1/20:1, and d18:1/20:0, d18:1/22:0, and

d18:1/24:0(Figure2.6b).Generally,allsphingomyelinspeciesshowedadownwardtrendinthe

VPS26a+/-mice,andinfacttheentireclassofsphingomyelinswasreduced(Figure2.6a),atrend

whichdidnotreachsignificanceinyoungercohorts.

37

Figure2.6AgeexacerbatessphingomyelinprofileinVPS26a+/-CSFandextendseffectstootherlipidclasses.

Inatargetedinvestigationofpositive-modeCSFlipids,threeentirelipidclassesaresignificantly

reduced: sphingomyelins (SM), dihydrosphingomyelins (dhSM), and monohexosylceramides

(MhCer). Values are normalized to thewildtype and represent themean +/- SEM (n=3-5; *P-

value<0.05,**P-value<0.01);dhCer=dihidroceramide;Cer=ceramide;LacCer= lactosylceramide;

GalCer=galactosylceramide; PC=phosphatidylcholine; PCEe=phosphatidylcholine ether;

LPC=lysophosphatidylcholine;LPCe=lysophosphatidylcholineether.(b)Abreakdownofindividual

sphingomyelinspeciesrevealsreductionof five lipids intheVPS26a+/-CSF:SMd18:1/18:1,SM

38

d18:1/20:0, SM d18:1/20:1, SM d18:1/22:0, SM d18:1/24:0. Values are normalized to the

wildtypeandrepresentthemean+/-SEM(n=3-5;*P-value<0.05,**P-value<0.01).

Unexpectedly,alterationsinotherlipidclassesappearedinthe6-month-oldVPS26a+/-CSF;these

included reductions inbothdihydrosphingomyelins andmonohexosylceramides (Figure2.6a).

Taken together, these studies indicate progressive sphingolipid dysregulation in the VPS26a-

haploinsufficient brain, which can be detected via CSF reporters such as sphingomyelin

d18:1/18:1atthreemonths(andlikelyotherspeciesatlatertimepoints).

2.2.5 SphingomyelindysregulationintheVPS26a+/-brainisnon-neuronal

VPS26haploinsufficiencyresultsinaknockdownofthiscoreretromerproteininallcellsofthe

brain.AsVPS26aisexpressedinallbraincelltypes(Figure2.7a)(Zhang,Chenetal.2014),itis

difficult to speculate which cell type might be responsible for imparting these sphingoliid

changes.Inordertoexplorethisinvivo,wemadeuseofourVPS26aneuron-specificknockout

(via Camk2a-Cre recombinase). Since VPS26a deficiency in thismousemodel is restricted to

neurons,acomparisonofCSFlipidomicsfromthetwoVPS26amodelscanindicatewhetherthe

sphingomyelindysregulationisdrivenbyneuronalretromerdeficiencyorinsteadbyothercell

types. (We acknowledge this isn’t a perfect comparison since the VPS26a+/- model is

haploinsufficientforVPS26awhiletheVPS26aCamk2aknockoutmodelisfullydepletedofthe

geneinneurons.However,ifthesphingomyelinchangeswerecomingfromneurons,wewould

expectanevenmorepronouncedphenotypeintheCamk2aknockouts.)

39

Figure 2.7 Sphingomyelin alterations in VPS26a-haploinsufficient mice derive from non-neuronalorigins.

(a) VPS26a is expressed in many cell types in the brain (Zhang, Chen et al. 2014). A global

assessment of lipidomic changes reveals no changes in sphingomyelins (SM),

dihydrosphingomyelins(dhSMs),ormonohexosylceramides(MhCer)intheVPS26aCamk2amice,

asnotedat thisage in theVPS26a-haploinsufficientmice. (b)Additionally,no changes inany

sphingomyelinsareobservedintheVPS26aCamk2aCSF.Valuesarenormalizedtothewildtype

andrepresentthemean+/-SEM(n=4-6;*P<0.05).

40

Tothisend,wecollectedCSFfrom6-month-oldVPS26aCamk2aknockoutmiceandlittermate

controlsandrepeatedthelipidomicsexperiment.Acomparisonoflipidclassesrevealsadistinct

globalprofileintheCSFoftheseretromer-deficientmiceascomparedwiththeVPS26a+/-model.

Changesinsphingomyelins,dihydrosphingomyelins,andmonohexosylceramidesobservedinCSF

from6-month-oldVPS26a+/-micearenotsharedinthe6-month-oldVPS26aCamk2aknockouts

(Figure2.7b).Furthermore,abreakdownofthesphingomyelinclassrevealsnochangesinany

individual lipid species (Figure 2.7c).We therefore conclude that the sphingomyelin changes

observedinVPS26a+/-CSFaretheresultofretromerdeficiencyinanon-neuronalcelltype.

Anadditionalconclusionfromthiscomparisonisthatretromerdeficiencyindifferentcelltypes

ofthebraincanleadtodistinctCSFlipidprofiles,andthatcompleteknockdownofVPS26ain

neuronsimpartsnodetectablechangestotheCSFlipidome.

2.3 Discussion

Thestudyoflipidsinthecontextofpathologyhasgainedincreasingmomentumoverthepast

fewdecades,propelledbyadvancementsinmassspectrometrytechnology.Toourknowledge,

thisstudyrepresentsthefirstattemptinutilizingthistechnologytocharacterizethemouseCSF

lipidome,and togauge its feasibility in reportingonendosomaldysfunction in thebrain.The

overwhelmingsimilarityinthelipidprofileofmouseandhumanCSFsuggeststhistechniqueasa

fruitfulmeansofbiomarkerdiscovery.

41

DespitethesmallvolumeofmouseCSFavailableformassspectrometry,172lipidspecieswere

detectedandquantifiedinourinitialscreenusingpooledreplicatesof30uLeach.This isover

50%ofthenumberdetectedin300uLofhumanCSF(318species)usingthesametechnology.It

is worth mentioning that pooling CSF from mice of the same genotype (30uL volume) was

favorable over analyzing individual samples (7uL volume) in maximizing the number of

quantifiable lipids (by 20% in the case of sphingomyelins). In the end,we deemmurine CSF

lipidomicsafeasiblemeansofbiomarkerdetectioninatleasthalfofthehumanCSFlipidome.

In this study,we conclude that VPS26a haploinsufficiency alters the sphingomyelin profile in

murine CSF and that the profile intensifies with age, extending into other lipid classes

(dihydrosphingomyelinsandmonohexosylceramides).Themostconsistentofthesechangesis

sphingomyelinisSMd18:1/18:1,whichwasdepictedinallcohortsfromeachagegroup.

It is interestingtonote,however,thatallofthesechangesseemtoderiveentirelyfromnon-

neuronalcells.Infact,completedeletionofVPS26ainneurons(viaCamk2a-Cre)resultedinno

significantalterationsintheCSF.Wewonderwhetherthismightindicateamoreprominentrole

fortheVPS26a-containingassemblyinaglialcelltype,atleastinitseffectsontheextracellular

lipidome.

Determining the source of these sphingomyelin changes in the brain would require the

generation of alternative cell type-specific knockdown models, or alternatively, in vitro

knockdownof individualcell types for lipidomicassessmentof theirsecretome.Furthermore,

42

studiesofCSFsphingomyelinsinADpatientshaveprovidedconflictingresults(thoughamajority

suggestelevationofthislipidclassinAD)(Kosicek,Zetterbergetal.2012,Fonteh,Ormsethetal.

2015),renderingthedevelopmentofasphingomyelinbiomarkerapotentiallycomplicatedtask.

Asthisprojectprogressedsimultaneouslyalongmultiplefronts,weoptedtoinsteadfocusour

effortsonourparallelproteomicstudieswhichyieldedmorecompellingresults.

3 Chapter3:BrainExosomes

3.1 Introduction

3.1.1 Exosomebiology

Exosomesaresmall(30-150nm),secreted,membrane-boundproductsoftheendosomalsystem.

They can travel via bodily fluids within and between organs, carrying with them preserved

material fromtheiroriginatingcells (Costa-Silva,Aielloetal.2015,Hoshino,Costa-Silvaetal.

2015,Zhang,Zhangetal.2015),andarethoughttobefairlyresistanttocirculatingproteasesin

biologicalfluids.Assuch,theyhavebeenincreasinglystudiedinthepastdecadefortheirability

toreportondiseasestates(Goetzl,Boxeretal.2015,Abner,Jichaetal.2016,Goetzl,Mustapic

etal.2016).Morerecently,theyhaveevenbeenimplicatedinvariouspathologies,asprimers

formetastasisofprimarytumorsandvehiclesforthespreadofproteinaggregatesinAD(Perez-

Gonzalez,Gauthieretal.2012,Asai,Ikezuetal.2015,Xiao,Zhangetal.2017).

3.1.2 Exosomesasreportersofendosomaldysfunction

43

Fourkeyobservationssupportthehypothesisthatretromer-mediatedendosomaldysfunctionin

themurinebrainmightalterthecontentofsecretedexosomes.First,ithasbeendemonstrated

byseveralindependentstudiesthatendolysosomaldysfunctioncanaltertheproteinand/orlipid

profileofexosomes(Yuyama,Yamamotoetal.2008,Strauss,Goebeletal.2010,Alvarez-Erviti,

Seowetal.2011).Second,retromerdysfunctionspecificallyhasbeenshowntoaltertheprotein

contentofexosomesinvitro(Sullivan,Jayetal.2011).Third,proteinswhichcomprisethecargo

recognitioncoreofretromerhavebeendetectedinexosomesofbodilyfluids(Reinhardt,Sacco

et al. 2013). And fourth, a recent study of neuronally-derived exosomes from AD patients

revealedchangesinagroupofendolysosomalproteins(Goetzl,Boxeretal.2015).

3.1.3 Brainexosomebiomarkerstrategy

Importantly,tworecentadvancementsinexosomeisolationmakefeasiblethestudyofmurine

brainexosomesforbiomarkerdiscovery.First,atechniqueforisolatingbrain-derivedexosomes

inhumanserumwasdescribed(Goetzl,Boxeretal.2015,Goetzl,Boxeretal.2015,Abner,Jicha

etal.2016),providingthefirsteveracquisitionofbrain-specificexosomesinvivo.Andsecond,

the development of a protocol for isolating adult mouse brain exosomes (Perez-Gonzalez,

Gauthieretal.2012,Asai,Ikezuetal.2015)allowstheinterrogationofrodentmodelsofdisease,

reducingthetranslationalhurdlesandboostingtheexosomeyieldoverconventionalcellculture

studies.Thus,at least in theory,biomarkers identified inmurinebrainexosomescannowbe

developedforevaluationinhumanserum-circulatingexosomesofbrainorigin.

44

Asaresult,weendeavoredtocharacterizethebrainexosomesfromseveralofourretromer-

deficientmodels.InformedbyourlipidomicstudyintheVPS26a-haploinsufficientmice(where

wenotedamildCSFlipidomicphenotypeofnon-neuronalorigin),weoptedtoperformourwork

incompleteretromerknockoutmodels:theVPS35fl/flCamk2a-CreKOandtheVPS26bknockout.

3.2 Results

3.2.1 Isolationofbrain-derivedexosomesfromwildtypemice:validationoftechnique

Exosomeisolationfromrodentbrainsisnotatrivialtask(Perez-Gonzalez,Gauthieretal.2012).

Itincludesmultipleroundsofultracentrifugations,andan18-houriodixanol(optiprep)gradient

separation(Figure3.1a).Weinitiallysoughttovalidatethisprotocolusingpreviouslyestablished

biochemicalandphysicalpropertiesofexosomes(Figure3.1b-d).Inapilotofwildtypemice,we

determined that gradient fractions 4-7 are enriched in exosomalmarkers alix and flotillin-1,

contain 30-150nm vesicles of exosomal morphology, and float between 1.07 and 1.11g/mL,

similartothosereportedpreviouslyforiodixanolgradientisolations(Collino,Pomattoetal.2017,

Klingeborn,Dismukeetal.2017).Assuch,weutilizedtheseexosome-enrichedfractionsforall

downstreamanalysesdescribedherein.

45

Figure3.1Isolationofhighqualityexosomesfromadultmurinebrain.

(a)Exosomeisolationfrombraintissuerequiresmultiplesteps.(b)Exosome-enrichedfractions

are positive for markers alix and flotillin-1 with limited contamination from other cellular

compartments(golgiasmeasuredbyGM130andERasmeasuredbycalnexin).(c)Thesefractions

floatatadensityof1.07-1.11g/mL,similar to thatpreviously reported for iodixanolgradients

46

(Collino,Pomattoetal.2017,Klingeborn,Dismukeetal.2017).(d)Electronmicroscopyreveals

vesicles30-150nmindiameterdisplayingtheconventionalexosomemorphology.

3.2.2 Retromercoreproteinsarepresentinbrainexosomefractions

Several groups have reported the presence of retromer cargo recognition core proteins in

exosomes of various tissues or fluids (Gonzales, Pisitkun et al. 2009, Demory Beckler,

Higginbotham et al. 2013, Skogberg, Gudmundsdottir et al. 2013). Therefore, we wondered

whetherwecoulddetecttheseretromerproteinsinbrain-derivedexosomesofadultmice.Using

the technique described above,we isolated exosomes from the brains ofwildtypemice and

probedforretromercoreproteinsVPS35andVPS26a(Figure3.2a).Infact,bothoftheseproteins

were present preferentially in the exosome-enriched fractions. In a separate study, we also

detectedretromercoreproteinVPS26aonintraluminalvesicles(ILVs)ofmulti-vesicularbodies

(MVBs)inneuronalcultures(Figure3.2b);asMVBscanultimatelygiverisetoexosomesviafusion

withtheplasmamembrane,wepresentthisobservationinsupportofthenotionthatretromer

cargoproteinscanassociatewithexosomesinthebrain.

47

Figure3.2Retromercargorecognitioncoreproteinsarepresentinbrainexosomefractions.

(a)CargorecognitioncoreproteinsVPS35andVPS26aaredetectedinexosomefractionsisolated

fromwildtypemousebrain.(b)RetromercoreproteinVPS26aisfoundonintraluminalvesicles

(ILVs)ofamultivesicularbody(MVB)inwildtypemouseneurons.

3.2.3 Brainexosomesderivedfromretromer-deficientmicehavereducedretromerprotein

levels

Wenextaskedwhetherthelevelsofretromerproteinsinbrain-derivedexosomeswouldreflect

thoseinthebraintissue,andassuch,provideameansofreportingonretromerdeficiencyinthe

brain.Toinvestigate,weisolatedbrainexosomesfromtworetromerknockoutmodels(VPS26b

knockoutandVPS35Camk2aknockout)andcomparedthelevelsofretromercargoproteinswith

thoseinexosomesofthecontrols(Figure3.3).

Micelackingneuronal-enrichedcoreproteinVPS26bexhibitadeficiencyofcoreproteinsVPS35

andVPS29inthebrain,whilelevelsofitsparalogVPS26aarelargelyunchanged,attributedto

theirnonreduntantinclusionindistinctretromerassemblies(Figure3.3a).Brainexosomesfrom

thesemicerevealanoticeablysimilarpatternamongretromerproteinlevels(Figure3.3a).

48

Figure3.3Murinebrainexosomesreflectretromerdeficiencyinthebrain.

(a)MicebearinghomozygousdeletionofcoreproteinVPS26bhavereducedlevelsofcoreproteins

VPS35,VPS26a,andVPS29butnochangesinVPS26ainbothbraintissueandinbrain-derived

exosomes. (b)Micewith neuron-specific deletion of VPS35 (via Camk2a-Cre) showdecreased

levels of other core proteins in both the brain and in brain-derived exosomes. Values are

normalized to the wildtype and represent the mean +/- SEM (n=3-9; *P-value<0.05, **P-

value<0.01,***P-value<0.001,****P-value<0.0001;withbrainlysatecontributionfromSabrina

Simoes).

KnockoutoftheVPS35viaCamk2a-Crerecombinasecausesalossofallretromercoreproteins

inthebrain(Figure3.3b).Ofthecoreproteinsmeasuredinexosomesfromthesemice(VPS35,

VPS26b,andVPS29),wefoundasimilarpatternofdeficiency(Figure3.3b).Collectively,these

49

datasuggestthatbrain-derivedexosomefractions,whichcontainretromercoreproteins,may

offerutilityasareporterofretromerlevelsinthebrain.

3.2.4 Brainexosomesderivedfromretromerknockoutmiceareenrichedinb-CTFs

Whileabiomarkerofretromerdeficiencymayproveusefulforasubsetofpatients,itispossible

thatAD-relatedretromerdysfunctioncanresultfromgeneticmutationswhichhavenoeffecton

overall retromer levels. Therefore, we sought to investigate other functional readouts of

retromerdeficiencywhichmightbereleasedinassociationwithexosomes.

APP, as described above, is a transmembrane protein which can be cleaved by several

membrane-associated proteases in various cellular compartments. The resulting C-terminal

fragments(CTFs)havebeenfoundinexosomesfromculturedcells(Sullivan,Jayetal.2011),as

well as brain-derived exosomes (Perez-Gonzalez, Gauthier et al. 2012). Retromer deficiency

increasesretentionofAPPintheendosome(Bhalla,Vetanovetzetal.2012),where it ismost

likelytobecleavedbyBACE1(SmallandGandy2006),andendosomalmembrane(bearingthe

BACE1-cleaved CTF) serves as starting material for exosomal biogenesis. Therefore, we

hypothesizedthatexosomesfromretromer-deficientbrainswouldbeenrichedintheseBACE1-

cleavedCTFs(b-CTFs).Infact,theelevationofCTFsinexosomeshaspreviouslybeendescribed

asaresultofretromerdeficiencyinHEK-293cells(Sullivan,Jayetal.2011).

To this end,we isolatedexosomes from the twomodels describedaboveand theirwildtype

littermatesandprobedthesampleswithanantibodydirectedtowardanepitopewithintheb-

50

CTF region of APP (M3.2). Interestingly, brain exosomes fromboth the neuron-enriched and

neuron-selectiveknockouts(VPS26bknockoutandVPS35Camk2aknockout)revealedastriking

elevationinthelevelsofb-CTFs(Figure3.4a).Wewereunabletodetectfull-lengthAPPinthe

brainexosomesofeitherthewildtypeorknockoutmice,whichcorroboratespriorreportsthat

full-lengthAPPisscarcelyfoundinexosomesascomparedwithitscleavedC-terminalfragments

(Perez-Gonzalez,Gauthieretal.2012).AcomparisonwithhippocampallysatefromtheVPS26b

knockouts(Figure3.4b)indicatesthatthiseffectisconcentratedintheendosomalcompartments

whichgiverisetotheseexosomes.Takentogether,thesefindingsdemonstratethatincreased

amyloidogenic processing of APP in the endosome of neuronal retromer knockouts can be

detectedinbrain-derivedexosomesviaenrichmentofb-CTFs.

51

Figure3.4Retromerdeficiencyinvivopromotesb-CTFaccumulationinexosomes.

(a)FullknockoutofVPS26bandneuronal-selectiveknockoutofVPS35(viaCamk2aCre)increase

theaccumulationofb-CTFsdetectedinpurifiedbrainexosomes.Valuesarefirstnormalizedto

ponceauandthenthewildtypeaverageandrepresentthemean+/-SEM(n=8-9;*P-value<0.05,

**P-value<0.01).(b)HippocampaltissuelysatesfromVPS26bknockoutmiceshowonlyamodest

trendtowardincreaseinb-CTFs(n=4;tissuelysatecontributedbySabrinaSimoes).

52

3.2.5 Brain exosomes from retromer knockout mice have decreased levels of exosomal

markersalixandflotillin-1

Inthecourseofourbrainexosomestudies,wehappeneduponanentirelyunexpectedfinding:

that brain exosomes from retromer knockoutmice contain lower levels of two conventional

exosomemarkersalixandflotillin-1(Figure3.5).Thoughthistrenddidnotreachsignificancein

theVPS35Camk2aknockoutmice(perhapsduetohighvariabilitybetweensamples),itwasclear

intheVPS26bknockoutmodel,witharoughly50%reductionasmeasuredbyimmunoblot.

Figure 3.5 Brain exosomes fromVPS26b knockoutmice contain reduced levels of canonicalexosomemarkers

(a)BrainexosomesisolatedfromVPS26bknockoutmicehavemarkedlyreducedlevelsofexosome

markersalixandflotillin-1asmeasuredbywesternblot.(b)Atrendtowarddecreaseofalixand

flotillin-1,thoughnotsignificant,isalsoobservedintheVPS35Camk2amodel.Valuesarefirst

normalizedtoponceauandthenthewildtypeaverageandrepresentthemean+/-SEM(n=8;*P-

value<0.05,**P-value<0.01).

As alix and flotillin are often used to confirm exosomal identity, wewonderedwhether the

decrease intheseproteins intheexosomal fractionssuggestchanges inexosomalbiogenesis,

andpossibly theirmorphology.Accordingly,weexamined thebrainexosomes fromretromer

53

knockout (VPS26b-/-) and control (VPS26b+/+)mice using electronmicroscopy. Though overall

exosomalmorphologyseemedtobefairlyconsistentbetweenthetwogroups,wenotedwhat

seemedtobeashifttowardsmaller(<50nMindiameter)exosomesintheretromerknockout

fractionsascomparedwiththecontrol(Figure3.6).Thisdescriptiveobservationwasnottested

statisticallyusingbiologicalreplicates.However,iftrue,thisfindingwouldfurthersuggestthat

retromer-dependentendosomaldysfunctioncanalter thebiogenicpathwaysofexosomes, in

additiontoregulatingtheircontent.

Figure3.6DescriptiveelectronmicroscopyrevealsenrichmentofsmallervesiclesinVPS26b-/-brainexosomes.

54

A descriptive analysis (n=1) of exosome-containing fractions from VPS26-/- vs control brains

revealsaperceivedenrichmentofsmaller(<50nmvesicles,examplesindicatedwithblackarrows)

intheknockoutcondition.Statisticswerenotperformed.

3.3 Discussion

Ourstudiesconfirmthepresenceofcargorecognitioncoreproteinsintheexosome-containing

fractionsfrommurinebrainandrevealthattheirlevelsreflectthatofthecellsfromwhichthey

originate.Thus,atleastinspecificcasesofretromerdeficiency,theymaybesuitedtoreporton

overall retromer levels in the brain. We opted not to immediately pursue this finding for

biomarker development, however, given that AD-related endosomal dysfunctionmay not be

limitedtoareductionofretromerlevels.

Priorstudieshavedemonstratedthatendolysosomaldysfunctioncanalterthecompositionof

secretedexosomes(Yuyama,Yamamotoetal.2008,Strauss,Goebeletal.2010,Sullivan,Jayet

al.2011).Ourworkisamongthefirsttoshowconfirmationofthisphenomenoninvivoandthat

brain-derivedexosomesmightreportonbrain-specificendosomaldysfunction.Givenretromer’s

roleindivertingAPPfromamyloidogenicprocessingbyBACE1intheendosome(Small,Kentet

al.2005,SmallandGandy2006,Wen,Tangetal.2011,Bhalla,Vetanovetzetal.2012),itisn’t

surprisingthatretromerdeficiencypromotesaccumulationofb-CTFs,whichcanbeloadedonto

ILVsandsecretedasexosomesintotheextracellularspace(Perez-Gonzalez,Gauthieretal.2012).

Itisreasonabletopostulate,however,thatanycellulardefectresultinginincreasedAPP(orits

55

cleavageenzymeBACE1)intheneuronalendosomemightpromotetheincreasedaccumulation

oftheseamyloidogenicfragmentsinbrain-derivedexosomes.

Unfortunately,wewereunabletotranslatethisfindingtohumanpatients,aswefailedtodetect

b-CTFsinserumexosomesofbrainoriginviaimmunoblot.Thisresultcouldarisefromatleast

three different possibilities. First, this lack of detection could simply be due to inadequate

sensitivityoftheavailabletechnologies;inthiscase,thedevelopmentofamoresensitiveassay

(suchasasimoa(Childs-Disney,Wuetal.2007,Wu,Milutinovicetal.2015))forthedetectionof

b-CTFscouldbeuseful.Thesecondpossibility isthatexosome-associatedb-CTFsarenotfully

retainedinexosomesthatcrosstheblood-brainbarrier.Ascleavageenzymessuchasa-secretase

(ADAM10), BACE1, and g-secretase (nicastrin) have been detected in brain-derived exosome

fractions (Perez-Gonzalez,Gauthier et al. 2012), onemust consider that CTFs from thebrain

mightbeprogressivelyprocessedthroughouttheirtransportintothebloodstream.Andthird,it

remainsentirelypossiblethatexosomeswhichhouseb-CTFssimplydonotcrosstheblood-brain

barrier. While a growing number of studies have shown delivery of exosomes from the

bloodstream into the brain (Yang, Martin et al. 2015, Khalyfa, Gozal et al. 2017), few have

describedthereversetransport(Goetzl,Boxeretal.2015,Goetzl,Mustapicetal.2016),andthe

mechanisms for exosomal traverse of the blood-brain remain largely unclear. (Whether

exosomesaredeliveredacrossthisbarrierinacontent-specificmanneriscurrentlyunknownbut

shouldbeconsideredasonepotentialexplanation for the lackofb-CTFs incirculatingserum

exosomes.) If either of the latter two possibilities are true, then continued work for the

developmentofanexosome-basedbiomarkershouldbestrategizedtoovercomethechallenge

56

of transport into the bloodstream. Accordingly, collection and interrogation (e.g. via mass

spectrometry)ofneuronally-derivedexosomesalreadycirculating in thebloodstreamofmice

withneuron-specificretromerdeficiencycouldprovideanoptimalhavenfordiscovery.

Finally,wereportanunexpected finding: that retromerdeficiency invivomayalterexosome

biogenic pathways. This conclusion is supported by a decrease in exosomemarkers alix and

flotillin-1 in the exosome fractions from retromer-deficient mice and a potential (as yet

descriptive)shiftinexosomesize.RecentstudiesbyotherlabshavedemonstratedthatAlixmay

be functionally regulated by Arf6 (Ghossoub, Lembo et al. 2014), which can also regulate

retromer (Marquer,Tianetal.2016).Thoughthedirectionalityof theserelationshipsdoesn’t

clarify the observed transcriptional effects of retromer deficiency on alix in our studies, it

supports the notion that these processes of retromer-dependent endosomal trafficking and

exosomebiogenesisaretightlylinkedinthemammalianbrain.Interestingly,workbyoneofour

collaboratorsrevealedthatneuronally-derivedexosomesfromADpatientsaresmallerandare

deficientincanonicalexosomemarkers(inthiscaseTSG101andCD81)thanthosefromcontrols

(Tran2017).

4 Chapter4:CSFProteomics

4.1 Introduction

4.1.1 Endosomaldysfunctionandthesecretedproteome

57

Itiswelldocumentedthatdefectsinendolysosomalfunctioncanalterthesecretedproteome

(Rojas, van Vlijmen et al. 2008, Tholen, Biniossek et al. 2011, Tholen, Biniossek et al. 2014).

Moreover, recent work has demonstrated that retromer-dependent endosomal dysfunction

resultsinanincreasedsecretionofAb(Bhalla,Vetanovetzetal.2012)andcathepsinD(Rojas,

vanVlijmenetal.2008).BothofthesemoleculesmigrateintotheCSF,andinfactaccumulatein

ADpatientsalongwithotherkeycomponentsoftheendolysosomalsystem(Schwagerl,Mohan

etal.1995,Armstrong,Mattssonetal.2014). Takentogether, theseobservations leadus to

hypothesize that retromer deficiency in the rodent brain will induce measurable proteomic

alterationsintheCSF.

4.1.2 CSFProteomicsbiomarkerstrategy

Asmentionedpreviously,CSFcollectioninrodentsisahighlytechnicalprocedurewhichyields

minimalfluid(roughly5uL).Nonetheless,severalgroupshavesuccessfullyperformedproteomics

onmurineCSF,detectingover700proteinsinaslittleas2uLoffluidperreplicate(Smith,Angel

etal.2014,Dislich,Wohlrabetal.2015).Wethereforeoptedtousesuchanunbiasedapproach

to maximize our biomarker discovery efforts in favor of hypothesis-driven immunoblotting

(which isempiricallybiased, reliesonavailabilityof reactiveantibodies,andcanrequireeven

moreCSFperreplicatetodetectsignal).

CSFisrelativelylowinproteincontent(0.2-0.8mg/mLinhumans(Rozek,Ricardo-Dukelowetal.

2007,ShoresandKnapp2007))ascomparedwithotherbiofluidssuchasserum(60-80mg/mL).

Furthermore,CSFproteinsspanawidedynamicrange,upto12ordersofmagnitudeinhumans

58

(Huhmer, Biringer et al. 2006, Schutzer, Liu et al. 2010),with albumin and immunoglobulins

occupyingalargemajorityofthetotal(YuanandDesiderio2005).Therefore,asinthelipidomics

study,we selecteda strategy that involved thepoolingof individual samples in thehopesof

increasingthenumberofdetectableproteinsavailableforbiomarkerdiscovery.

Ourstudydesigninvolvedthecollectionofhighquality(bloodcontamination-free)CSFsamples

tobepooledintobiologicalreplicatesof30uLeach.ThesewererunviaLCMS/MSusingeithera

labeled or unlabeled approach, and any resulting hits from the screen were followed up in

biochemicalanalysesofnewcohorts.

WeelectedtobeginwiththeVPS26a-haploinsufficientmodelaswefeltthelevelsofretromer

knockdownwouldbemorereflectiveofthatseeninADpatientsandasthismodeldoesnotlimit

retromerdysfunctiontoneurons.We laterperformedasimilarstudyontheVPS35knockout

model,whichexhibitsfullneuronaldeletion(viaCamk2a-Cre)ofthislinchpintoretromer’score.

Resultsfromeachstudyareincludedinthissection.

4.2 Results

4.2.1 CSF from VPS26a+/- mice reveals no apparent protein alterations as measured via

LC/MS-MS

HighqualityCSFwascollectedfrom3-month-oldVPS26a+/-miceandwildtypelittermatesand

pooledinto4biologicalreplicatespergenotype.Thesesampleswereeachlabeledwithdistinct

isobaricTMTtagsandrunviaLC/MS-MSalongwithacompositereferencesample,whichwas

59

used for normalization (Figure 4.1). This strategy interprets signal from each isobaric tag as

relativesamplecontribution(i.e.relativequantitation)fortheproteinsidentified.

Figure4.1OverviewoflabeledLC-MS/MSapproachforproteomicanalysisofVPS26a+/-CSF.

CSFwascollectedfrom3-month-oldVPS26+/-malesandlittermatecontrolsandassessedforblood

contaminationviahemoglobinELISA.Contamination-freesamplesweresubsequentlypooledinto

biologicalreplicatesof30uLeach(n=4pergenotype).Proteinsweredigestedandlabeledwitha

distinct isobaricTMTtag,combined,andfractionatedbypHtoreducesamplecomplexity.LC-

MS/MSwassubsequentlyperformedonthefractionatedsamplesfollowedbystatisticalanalyses.

Using this labeled approach, 1215 quantifiable proteins were positively identified in the

combinedCSF.Histogramsofthelogratios(sample/reference)forallidentifiedproteinsreveal

fairlynormaldistributionsamongthebiologicalreplicates(Figure4.2).

60

Figure4.2Qualityassessmentrevealsnormallydistributedproteinquantitationforallsamples.

Foreachsample,relativequantitationforeachprotein(measuredassample/referencesignal)

wasperformedandreportedlogarithmically(logratio).Ahistogramofthese1215logratiosis

normallydistributedforeachofthefourVPS26a+/-(KD1-4)andVPS26a+/+(WT1-4)samples.

Surprisingly, however, of all 1215 proteins detected, none were significantly altered in the

VPS26a+/-CSFaftercontrollingformultipletesting.Infact,acorrelationplotofthelogratiosfor

the retromer-deficient (KD)CSFversuscontrol (WT) revealsanR-squaredof0.95 (Figure4.3)

indicatingaveryhighlevelofcorrelationofCSFproteinsbetweenthetwogenotypes.

61

Figure4.3GlobalCSFprofileinVPS26a+/-micelargelyresemblesthatoflittermatecontrols.

AcomparisonofmeanlogratiosbetweentheVPS26a+/-(KD)andlittermatecontrols(WT)reveals

ahighlevelofcorrelationbetweenthetwosamples.

Wewonderedwhetherperhapsthestatistical‘penalty’formultiplecomparisonwasmaskingreal

butmodestproteinchangeswithinourdataset.Therefore,welooseneddiscoveryrestrictions

andperformedaregressionanalysistorevealthetopstatisticalcandidates(thoughnonemet

the multiple testing-corrected significance threshold). Among the handful of statistical

candidateswere severalwhich have been previously linked to the endosomal system. These

includedfolatereceptor1(FOLR1),GDNFfamilyreceptoralpha1(GFRA1),profilin(PFN1),and

platelet-activatingfactoracetylhydrolase1b3(PAFAH1B3).GFRA1isknowntobetraffickedby

SORL1,a retromer cargo,outof theendosome (Glerup, Lumeetal. 2013). FOLR1undergoes

endosomal recyclingback to thecell surface, though theprecisemechanismremainsunclear

(Wibowo,Singhetal.2013).PFN1isinvolvedinactinremodeling,oneofthefivefunctionsofthe

62

retromerassembly,andinfluencesCD36recyclingwhichhasbeenshowntodependonretromer

(Romeo,Moultonetal.2007,Harbour,Breusegemetal.2010,Lucin,O'Brienetal.2013).And

PAFAH1B3playsacriticalroleinendosomaltubuleformation,anotherfunctionoftheretromer

assembly (Bechler, Doody et al. 2011). All of these proteins were elevated in the VPS26a

knockdownCSFexceptforPAFAH1B3.Asthese‘hits’wereonlyidentifiedthroughalooseningof

thestatisticalparameters,itwasimperativethatweassesstheirCSFlevelsinanewcohortof

VPS26a+/-mice.Wethusturnedtoimmunoblotinpursuitofantibodieswithreactivityinmurine

CSF.

Oftheseproteins,onlyPAFAH1B3hadacommerciallyavailableantibodywhichshowedreliable

reactivityinmurineCSF.InitialoptimizationstudiesinmouseCSFindicatedthatlargevolumesof

CSF(20uLperreplicate)wererequiredtodetectsignal.Wethereforeperformedawesternblot

usingpooledCSFfromVPS26a+/-miceandlittermatecontrolstoassessPAFAH1B3levelsinthe

CSF.Due to the largevolumerequiredper lane,only twobiological replicateswereusedper

condition.Whilewerealizethissmallsamplesizeisinsufficientformostbiochemicalassays,our

goal was the identification of a distinctive biomarker of endosomal dysfunction. Thus, we

reasoned that this follow-up validation (of pooled samples which should minimize variance

among sampleswithin each genotype) should properly informon the purporteddecrease of

PAFAH1B3intheCSFofthesemice.Resultsoftheimmunoblotrevealed,however,thatthelevels

of this protein are unchanged in the CSF of VPS26a+/-mice (Figure 4.4), rendering the initial

statisticalassessmentofthisprotein(asunchanged)correct.

63

Figure4.4PAFAH1B3isunchangedinVPS26a+/-CSF.

ImmunoblotofpooledCSFrevealsnochangeinPAFAH1B3levels inretromerknockdownmice

(n=2 due to limited availability of CSF). Values are first normalized to ponceau and then the

wildtypeaverageandrepresentthemean+/-SEM.

We ultimately conclude that VPS26a haploinsufficiency results in a mild CSF proteomic

phenotype,withnodetectablechangesvisLC-MS/MSattheageofthreemonths.Perhaps,asis

thecasefortheCSFlipidome,agemightexacerbateproteomicchanges,allowingdetectionata

latertimepoint(e.g.sixmonths),orperhapsacompleteknockoutisbettersuitedforsuchhigh-

throughputbiomarkerdiscovery.

4.2.2 MassspectrometryrevealsenrichmentofBACE1substrateN-terminalfragmentsand

totaltauinCSFofVPS35Camk2aknockoutmice

4.2.2.1MassspectrometryofVPS35Camk2amiceprovidesdeepestcharacterizationofmurine

CSFproteometodate

64

Informedby theabove-mentionedwork inourVPS26a+/-model,weopted to repeat theCSF

proteomicscreenusinganeuron-specificretromerknockoutmodelatsixmonthsofagehoping

thatthesetwoadjustmentswouldhelpexacerbateaproteomicphenotypewemightbeableto

detect.

Inthisinstance,weemployedasimilarpoolingstrategy,butowingtothelowlevelsofproteinin

CSF collected from thismodel (0.10-0.16µg/mLas comparedwith0.68-0.8µg/mL in theprior

model),weoptedforanunlabeledapproach(Figure4.5),whichrequiressequentialrunsofeach

biologicalreplicate.

Figure4.5OverviewoflabeledLC-MS/MSapproachforproteomicanalysisofVPS35Camk2αknockoutCSF.

CSFwascollectedfrom6-month-oldVPS35knockoutmalesandlittermatecontrolsandassessed

forbloodcontaminationviahemoglobinELISA.Contamination-freesamplesweresubsequently

65

pooledintobiologicalreplicatesof30uLeach(n=4-5pergenotype).Proteinsweredigestedand

labeledwithadistinct isobaricTMT tag, combined,and fractionatedbypH to reduce sample

complexity. LC-MS/MSwas subsequently performed on the fractionated samples followed by

statisticalanalyses.

Usingthisstrategy,1505proteinswerepositivelyidentifiedintheCSFby1uniquepeptide(1048

proteinsby2uniquepeptides).Todate,priorproteomicsstudiesofmurineCSF(usingasimilar

unlabeledapproachonsmallervolumes)havedetectedasmanyas817proteinsbyoneunique

peptideand715withatleasttwouniquepeptidesequences(Smith,Angeletal.2014,Dislich,

Wohlrabetal.2015)(Figure4.6),renderingthisyetthedeepestcharacterizationtodateofthe

murineCSFproteome.

Figure4.6ComparisonwithpriormurineCSFproteomicstudies.

66

Usingapooled,unlabeledapproachintheVPS35Camk2aknockoutmodel,weidentifiednearly

doubletheproteinsreportedpreviouslywithoneuniquepeptide,androughlyonethirdmorethan

previouslyreportedwithtwouniquepeptides(Smith,Angeletal.2014,Dislich,Wohlrabetal.

2015).

Technicalreplicateswerehighlycorrelated(Figure4.7a)aswerethemeanscoresforbiological

replicates of the same genotype (Figure 4.7b), indicating robustness of technique. Quality

assessmentsperformedbyourProteomicCoreidentifiedoneoutlierfromeachgenotypewhich

wasremovedpriortodownstreamanalyses.

Figure4.7Qualityassessmentrevealshighlevelsofcorrelationbetweenreplicates.

67

(a) A comparison of label-free quantification (LFQ) intensities for each protein reveal a high

correlationbetweentechnicalreplicates.(b)AcomparisonofLFQintensitiesforeachproteinin

biological replicates (represented as the mean of technical duplicates) demonstrates high

correlationbetweenbiologicalreplicates.

InourquestforaCSFbiomarkerofretromerdysfunction,wedecidedtoemploytwoanalytical

strategies: 1) a standard parametric analysis of proteins detected in samples from both

genotypes,and2)anonparametricanalysisinwhichweaskedwhichproteinsarepresentinone

condition but not the other (as such a categorical distinction might provide even greater

biomarkerpotential).

4.2.2.2Parametric analysis reveals changes in BACE1 substrate N-terminal fragments and

apolipoproteins

Ofthe1505proteinsidentifiedinthisstudy,52weresignificantlyalteredintheVPS35knockout

CSFasdeterminedbystandardANOVA(Table4.1).Withinthislist,aclassofelevatedproteins

wereofparticularinterest:BACE1substrates(CHL1,APLP1,andAPLP2,andAPP)(Figure4.8a).

PeptidesfromtheseBACE1substratesmappedontotheN-terminalfragments(NTFs),suggesting

mechanisticconvergenceuponBACE1cleavage(Figure4.8b),whichliberatestheNTFintothe

endosomallumen(andultimatelytheextracellularspace).

68

Table4.1ParametricHitsofVPS35Camk2aCSFProteomics.

Accession Symbol p-value q-value Direction P40142 Tkt 6.26E-08 0.0001 ↓ Q06335 Aplp2 2.83E-07 0.0002 ↑ P17182 Eno1 9.96E-07 0.0003 ↓ P17563 Selenbp1 1.12E-06 0.0003 ↓ P70296 Pebp1 1.14E-06 0.0003 ↓ Q06890 Clu 1.11E-06 0.0003 ↑ G3UYU6 Ptprd 2.47E-06 0.0005 ↑ O70362 Gpld1 3.30E-06 0.0005 ↑ P01029 C4b 2.79E-06 0.0005 ↑ P06909 Cfh 4.44E-06 0.0005 ↑ P08905 Lyz2 4.10E-06 0.0005 ↑ P09411 Pgk1 3.31E-06 0.0005 ↓ P70232 Chl1 2.69E-06 0.0005 ↑ P45376 Akr1b1 5.86E-06 0.0006 ↓ P63054 Pcp4 7.80E-06 0.0008 ↑ Q8VCM7 Fgg 7.99E-06 0.0008 ↓ A0A0R4J0X7 Gapdhs 1.05E-05 0.0009 ↓ H7BX99 F2 1.16E-05 0.0009 ↑ P08226 Apoe 1.31E-05 0.0009 ↑ P32848 Pvalb 1.00E-05 0.0009 ↓ Q9D0F9 Pgm1 1.32E-05 0.0009 ↓ P55065 Pltp 1.48E-05 0.001 ↑ A0A087WR50 Fn1 1.84E-05 0.0011 ↑ P08249 Mdh2 1.78E-05 0.0011 ↓ P97290 Serping1 1.63E-05 0.0011 ↑ Q61361 Bcan 1.74E-05 0.0011 ↑ P01887 B2m 2.25E-05 0.0012 ↑ P13020 Gsn 2.10E-05 0.0012 ↑ Q9CPU0 Glo1 2.35E-05 0.0012 ↓ P15626 Gstm2 3.20E-05 0.0016 ↓ P17183 Eno2 3.34E-05 0.0016 ↓ Q62000 Ogn 3.19E-05 0.0016 ↑ Q640N1 Aebp1 4.04E-05 0.0018 ↑ A0A075B664 Iglv2 4.42E-05 0.0019 ↓ Q03157 Aplp1 4.36E-05 0.0019 ↑ Q8R480 Nup85 5.36E-05 0.0022 ↓ P50247 Ahcy 6.07E-05 0.0025 ↓ E9QPX1 Col18a1 6.45E-05 0.0026 ↑ P21550 Eno3 7.23E-05 0.0028 ↓ A0A0R4J107 Apeh 7.85E-05 0.003 ↓ P51910 Apod 8.21E-05 0.003 ↑ P00920 Ca2 8.89E-05 0.0032 ↓ P02802 Mt1 9.13E-05 0.0032 ↓ P29699 Ahsg 9.68E-05 0.0033 ↑ Q9DCD0 Pgd 9.75E-05 0.0033 ↓ P12023 App 1.09E-04 0.0036 ↑ Q04447 Ckb 1.16E-04 0.0037 ↓ Q5NC80 Nme1 1.24E-04 0.0039 ↓ P02089 Hbb-b2 1.31E-04 0.004 ↓ P12960 Cntn1 1.32E-04 0.004 ↑ P14152 Mdh1 1.35E-04 0.004 ↓ Q62165 Dag1 1.39E-04 0.004 ↑

69

Figure4.8BACE1substrateNTFsareelevatedinCSFfromVPS35Camk2αknockoutmice.

(a)Mass spectrometryofVPS35Camk2a knockoutCSF reveals significant increases inBACE1

substratesAPLP1,APLP2,CHL1,andAPP.(b)PeptidesmaptotheN-terminalfragment(NTF)for

eachBACE1substratedetected.ValuesarereportedasLFQintensityandrepresentthemeanof

technicalduplicates.Barsrepresentmean+/-SD(n=3-4;*P-value<0.05**P<0.01,***P<0.001).

4.2.2.3NonparametricAnalysisUncoversChangesinCSFTotalTau

We then performed a nonparametric analysis, looking for proteins which present in one

conditionbutnottheother,assuchproteins(whichwouldbeoverlookedusingstandardANOVA

analyses) could carry high biomarker potential. We first employed this strategy looking for

proteins present in at least four technical replicates (of eight total) only the VPS35 Camk2a

knockoutCSFbutcompletelyabsentincontrols.Thisqueryresultedinalistof11proteins(Figure

4.9)spanningvariousfunctionalcategories.Interestingly,amongthesecategoricalhitswasMapt

70

(tau),aproteinwhichhasclearpathologictiestoAD(Grundke-Iqbal,Iqbaletal.1986,Bejanin,

Schonhautetal.2017)andisreliablyelevatedintheCSFofADpatients(Andreasen,Minthonet

al.2001).Wethenusedasimilarstrategytoidentifyproteinspresentinatleastfourtechnical

replicates of the control CSF but absent from all knockout samples. This yielded a list of 8

additionalnonparametrichits(Figure4.9).

Figure4.9NonparametricanalysisrevealspresenceofMapt(tau)inonlyretromerknockoutCSF.

InanonparametricanalysisoftheVPS35Camk2aCSFmassspectrometrydata,eightproteins

were detected in at least four technical replicates of the wildtype CSF and absent from all

knockoutreplicatesand11proteins inat least fourtechnical replicatesofcontrolsbutabsent

fromallwildtypereplicates.Notably,mapt(tau)ispresentinonlytheretromerknockoutCSF.

4.2.2.4Ingenuity Pathway Analysis (IPA) indicates APP, MAPT, and PSEN1 among the top

upstreamregulatorsofCSFchangesinVPS35neuronalknockoutmice

71

To further clarify the wealth of information gained from this CSF study (which includes 52

parametric hits and 19 nonparametric hits), we turned to ingenuity pathway analysis (IPA)

(Kramer, Green et al. 2014) to determine which genes might be ‘regulating’ the global CSF

changes in our VPS35 knockoutmodel. This software relies upon established transcriptional

interactions to predict upstreammolecular convergence of global changes in large datasets

(Kramer,Greenetal.2014).Ofthetopsixpotentialregulators,threeofthem(APP,MAPT,and

PSEN1)arepathologicallylinkedtoAD,andthetoptwo(APPandMAPT)areaberrantlyprocessed

proteinswhichcomprisethecurrentCSFdiagnosticprofile(Figure4.10).

Figure 4.10 Ingenuity Pathway Analysis (IPA) implicates AD-related proteins as upstreamregulatorsinVPS35knockoutCSF.

IPArevealsAPP,Mapt,andPSEN1amongthetopsixupstream‘regulators’ofthepotential843

identifiedinthisset.

4.2.3 BACE1SubstrateNTFsandTotalTauareconfirmedashits inanewcohortofVPS35

Camk2aknockoutmice

72

4.2.3.1BACE1substrateNTFsareconfirmedviaimmunoblotinvivoandinvitro

IntriguedbytheBACE1substratehits,wesoughttoconfirmtheirelevationinvivoinanewcohort

ofmiceusingimmunoblottechnique.Ofthislist(CHL1,APLP1,APLP2,andAPP),weidentified

availableantibodiesforbothCHL1andAPLP1(directedattheN-terminalregion)thatthatwere

reactiveinmouseCSF.BothoftheseproteinswereagainelevatedintheCSFofthenewcohort,

confirming the initialmass spectrometry results (Figure4.11a).Molecularweight comparison

withcorticallysates(viaimmunoblotmigration)suggeststhattheseareindeedtheNTFs,notthe

full-lengthprotein(datanotshown).Invitroworksupportedthesefindings,aslevelsofCHL1and

APLP1NTFswere increased in themedia fromVPS35fl/fl neurons infectedwith Cre.GFP virus

versusthatofthecontrol(Figure4.11b).

Figure4.11ElevationofBACE1substrateNTFsinVPS35knockoutsisconfirmedinvivoandinvitro.

73

(a)CSF fromanewcohortofVPS35Camk2a knockoutmiceand littermatecontrols revealsa

significantincreaseintheNTFsofBACE1substratesAPLP1andCHL1,confirmingthesehitsfrom

the initial proteomic screen. Bars represent mean +/- SD (n=6-8; *P<0.05). (b) Media from

culturedprimaryVPS35fl/flneuronsinfectedwithCre.GFP(VPS35+/+)orD-Cre.GFP(VPS35-/-)is

enrichedinNTFsfromBACE1substratesAPLP1andCHL1.Barsrepresentmean+/-SD(n=6-7from

2independentexperiments;*P<0.05).

4.2.3.2Single molecule array (simoa) confirms elevation of total tau in VPS35 Camk2a

knockoutCSF

DetectionoftauinmouseCSFhasbeenachallengingtask,aslevelsinCSFareroughly50,000-

foldlowerthanthatofbraintissue(Barten,Cadelinaetal.2011).Asaresult,toourknowledge,

noonehasmanagedtodetectthisproteininmurineCSFusingconventionalimmunoblotorELISA

inwildtypemice.Infact,effortsonourparttodetecttauinmurineusingbothimmunoblotand

MSDELISAwereunsuccessful.Therefore,weturnedtosinglemoleculearray(simoa),arecently

developedtechnologywhichboastsdetectiondowntothelevelofasinglemolecule.Usingthis

approach,wewereabletoreliablydetecttotaltauinaslittleas5uLofmurineCSF.Furthermore,

a comparison of VPS35 Camk2a-Cre knockoutmice and their littermate controls revealed a

strikingincreaseintheknockoutCSF(Figure4.12),confirmingthisnonparametrichitfromour

initialscreen.

74

Figure4.12TotaltauissignificantlyelevatedinVPS35Camk2αknockoutCSF.

In a new cohort ofVPS35Camk2a knockoutmice, CSF total tau ismarkedly increased in the

knockoutascomparedwithcontrol.Dotsrepresentrawvalues.Barsrepresentmean+/-SD(n=5-

9;*P<0.05,**P<0.01,***P<0.001).

WefurthervalidatedthisfindinginvitrothroughknockoutofVPS35(vialentiviraldeliveryofCre-

recombinase in VPS35fl/fl primary neurons). Here, a comparison with controls revealed an

elevation in secreted total tau from the VPS35 knockouts while levels of secreted lactate

dehydrogenase(LDH)wasunchanged(Figure4.13),indicatingthattheincreaseinextracellular

tauresultsfromanactivemechanismindependentofcelldeath.

75

Figure4.13TotaltauisenrichedinmediafromVPS35-/-neuroncultures.

VPS35fl/flprimaryneuronsweretreatedwithCre.GFP(VPS35-/-)orD-Cre.GFP(VPS35+/+).(a)Total

tauandLDHweremeasuredinthemediausingMSDtechnologyandpromegaCyto-toxassay,

respectively. Values were normalized to total protein and then to the control. Bar graphs

representmean +/- SEM (n=7, *P<0.05; **P<0.01; ***P<0.001). (b) Immunoblot of neuronal

culturelysatedemonstrateslentiviralknockdownofVPS35.Valuesarenormalizedtotheloading

controlandtothewildtypemean.Bargraphsrepresentmean+/-SEM(n=5-6,*P<0.05;**P<0.01;

***P<0.001;****P<0.0001).

4.2.4 TranslationtohumanspositionsCSFtotaltauasabiomarkerofendosomaldysfunction

inAD

4.2.4.1BACE1SubstrateNTFsarehighlycorrelatedinmouseandhumanCSF

During the courseofourmurineCSF validation studies,we stumbleduponan intriguingand

unanticipatedfinding:thatBACE1substrateNTFsAPLP1andCHL1arehighlycorrelatedinmouse

CSF(Figure4.14a).WethereforeturnedtohumanCSF(non-dementedcontrols)toseewhether

thisrelationshipheld(Figure4.14b).Usingasimilarimmunoblotapproach,wediscoveredthat

76

APLP1andCHL1NTFsarealsohighlycorrelatedintheCSFofhumansubjects.AstheseNTFsare

liberatedbythesamecleavageenzyme(BACE1),wereasonthissuggeststhatthelevelsofAPLP1

andCHL1NTFsintheCSFdependprimarilyontheactivityand/orlocalizationofBACE1.

Figure4.14NTFsofBACE1substratesarecorrelatedinmouseandhumanCSF.

CSF from mice (n=27) (a) and healthy control (n=39) CSF (b) reveals a striking correlation

(p<0.0001 for mouse and human) between the NTFs of APLP1 and CHL1 as measured by

immunoblot.Valuesarenormalizedtoalbuminandthegroupaveragebeforecombiningcohorts

fromdifferentblots.

4.2.4.2APLP1iscorrelatedtoAb42andtotaltauinhumanCSF

Next,wewonderedwhetherthelevelsoftheseBACE1substrateNTFswerecorrelatedwithother

AD-relatedproteins(Ab42,totaltau,andptau-181)intheCSF.Weexploredtheserelationships

inhumanCSFassimoaassayswithmousereactivityhadnotyetbeendevelopedforAb42and

ptau-181.Remarkably,analysisof38non-dementedcontrolsrevealedthatCSFlevelsofAPLP1

77

NTFsaresignificantlycorrelatedwithAb42,totaltauandptau-181(Figure4.15;Ab42:p<0.01;

totaltau:p<0.0001;ptau-181:p<0.05;n=38).

ThecorrelationwithAb42isperhapsnotentirelysurprisingasbothAb42andAPLP1NTFsare

bothdownstreamproductsofBACE1cleavage.Yet,thecorrelationsbetweenAPLP1NTFsand

bothtotaltauandptau-181areunexplainedbycurrentknowledge,althoughrecentinvitrowork

has shownthat retromerdeficiencypromotes tauphosphorylation (Young,Fongetal.2018).

TheseassociationsintheCSFsuggestthatthepathwayscontrollingsecretionofBACE1-cleaved

substratesandtaumaybelinkedintheplaque-freemammalianbrain.

Figure4.15APLP1NTFsarehighlycorrelatedwithAβ42andtotaltauinhumanCSF.

Astudyof38non-dementedhealthycontrolsrevealedsignificantcorrelationsbetweenCSFN-

APLP1andAb42(p<0.01),totaltau(p<0.0001),andptau-181(p<0.05).

4.2.4.3APLP1NTFslikelybindtoamyloidplaques

78

ThepathologicprogressionofADgenerallyinvolvesthedevelopmentofextensiveextracellular

plaques,whichhavebeenknowntosopupsecretedproteins,ultimatelyaffectingtheirdetection

intheCSF(Fagan,Mintunetal.2006).Todate,thepropertieswhichrenderproteinslikelyto

sticktoplaquesarenotfullyunderstood,thoughsomehaveattemptedtoaddclaritytothisissue

(Gozal,Chengetal.2006).Theultimateaimofourprojectisthetranslationoffindingstohuman

subjects in order to identify all individualswith endosomal dysfunction (both pre- and post-

plaque development). As such, it is imperative that levels of our biomarker of endosomal

dysfunctionnotbeconfoundedbythepresenceofplaques.

AqueryoftheliteratureturneduponereportthatinfactAPP-relatedproteinssuchasAPLP1

andAPLP2canaccumulateinneuriticplaquesincontrolandADbrains(McNamara,Ruffetal.

1998).WeturnedagaintohumanCSFtoseewhetherAPLP1NTFsmightbestickingtoplaques

inhuman subjects.WeobtainedCSF from5patientswithMCIwhowerebiomarker-positive

(Ab42>210 and Tau/Ab42<0.39) for AD (i.e. ‘plaque-positive’). Using this limited set, we

performedwasastandardt-testofCSFAPLP1inplaque-positiveADpatientsvsage-andgender-

matchedcontrols.TheunderlyingassumptionforthisquerywasthatifAPLP1isnotstickingto

plaques, the levelswouldbeeitherunchangedorelevated inplaque-positivepatients versus

controls.Whatwefoundinsteadwasamarkeddecreaseintheplaque-positivecohort(Figure

4.16), suggesting that like Ab42, APLP1 NTFsmight be binding plaques in the AD brain.We

followedupwithacorrelationanalysisofAPLP1andAb42inCSFfromthe5ADpatientswiththe

assumptionthatifAPLP1isnotaccumulatinginplaquesasisAb42,thiscorrelationshouldbreak.

Despitethe lowstatisticalpowerofsuchasmalldataset, thecorrelationbetweenAPLP1and

79

Ab42 remained quite strong (p=.003) in AD brains (data not shown). Taken together, these

observationssupportpriorfindingsandsuggestthatCSFlevelsofotherBACE1cleavageproducts

suchasAPLP1NTFsmaybeconfoundedbyplaques intheADbrain,thuscompromisingtheir

potentialforbiomarkerutility.

Figure4.16CSFLevelsofAPLP1NTFsarelowerinADpatients.

N-APLP1islowerintheCSFfromplaque-positiveADpatientsthaninage-andgender-matched

controls(n=5,22;*P<0.05).

4.3 Discussion

CSFcollectioninmiceisahighlytechnicalprocedure,oftenyieldingnomorethan5uLoffluidper

mouse. Nonetheless, other groups have successfully characterized this fluid via LC-MS/MS

(Smith,Angeletal.2014,Dislich,Wohlrabetal.2015),identifyingover700proteinsinaslittleas

2uL. As such, we sought out to uncover CSF protein biomarkers in two different models of

retromerdeficiencyusingasimilarstrategy.

80

4.3.1 VPS26a+/-CSFProteomics

ThatwefoundnoproteinchangesintheVPS26a-haploinsufficientCSFwassurprising,particularly

given the number of positively identified proteins (1215 in pooled replicates) available for

discovery. There could be several explanations for this. First, it is entirely possible that a

heterozygousmodelisinsufficienttogenerateaclearprotein-basedprofileintheCSF(oratleast

onestrongenoughtoovercomethemultiplecomparison‘penalties’insuchalargedataset).As

welearnedthroughourlipidomicsstudiesofthismodel,perhapsusingalatertimepointinthese

mice (e.g. six months instead of three—the age used for our VPS35 Camk2a model) would

increasethelikelihoodofdiscoverythroughanexacerbationofcellularpathology.

A second explanation for is that the VPS26a-containing assembly is less critical to trafficking

eventswhichalter the cellular secretome.Yet,weknow from in vitro studies thatdefects in

VPS26aincreasethesecretionofproteinssuchascathepsinD(Rojas,vanVlijmenetal.2008);

thus,weargueagainstsuchanexplanation.

4.3.2 VPS35Camk2aCSFProteomics

OurunbiasedCSFscreenofneuronal-selectiveVps35knockoutmiceyieldedacomplementof

proteins that most reliably accumulate in the CSF in a model of AD-associated endosomal

dysfunction.Aftervalidatingthemostrelevantproteins,weturnedtohumanCSFtoinvestigate

whethertheproposedrelationshipbetweentheseproteinsoccursinthehumanbrain.Themain

advantage of starting with a mouse model, besides greater experimental control, is that

endosomaldysfunctioncanberestrictedtocorticalneurons,obviatinganyconfusionofcelltype

81

specificity.Althoughthereisasignificantoverlapinendosomalmachinerybetweenmiceand

humans, cross-species differencesmight exist. Thus,while our positive findings are deemed

reliable,wecannotexcludetheimportanceofproteinsthatdidnotemergefromourprimary

mousescreen.

OurVPS35Camk2aknockoutCSFstudydemonstratesthedeepestprobeintothemurineCSF

proteome to date (by roughly 100% using a threshold of 1 unique peptide and 50%with a

threshold of 2 unique peptides)with an unlabeled approach. This featmay be attributed to

technicaldifferences(e.g.equipmentandhandling),butitislikelyatleastinpartaresultofthe

poolingstrategywedesigned,allowingalargervolumepersampleavailableforproteomicstudy

thanthoseusedinpriorstudies.

4.3.2.1BACE1SubstrateNTFsinMiceandHumans

ThefactthatAPPfragmentswereamongthehitsfromthisscreenprovidesadditionalunbiased

evidencelinkingretromertoAD.Morebroadly,anumberofpreviousstudies,includingaCSF

proteomic screen inBACE1KOmice,haveestablished thatAPP,APLP2,APLP1,andCHL1are

substrates cleavedbyBACE1 in theneuronal endosome, generating and secretingn-terminal

fragments of each. Our screen identified fragments of all of these proteins, fromwhichwe

concludethatretromer-dependentendosomaldysfunctionpromotestheiracceleratedcleavage

and secretion. All four of these proteins, including BACE1 itself, are type I transmembrane

proteins that traffic through theendosome. Retromer typically functions in trafficking these

typesoftransmembraneproteins,andindeed,thusfar,bothAPPandBACE1havebeenfoundto

82

betraffickedbyretromer,withretromerdeficiencyincreasingtheirresidenttimeinendosomal

membranes.Thus,theobservedaccumulationofthefourfragmentsmayreflectacombination

ofmistraffickedsubstrateandtheircleavingenzyme.WenotethatwedidnotdetectAβinthe

CSFofWTorKOmice,eventhoughretromerdeficiencydoescauseanincreaseinAβinmouse

brains,andinthemediaofculturedneurons.Weconsideritlikely,therefore,thatendogenous

murine Aβ is below detection of the methods used. Our human studies support this

interpretation, suggesting that in plaque-free individuals a marker of retromer-dependent

endosomaldysfunctionispositivelyassociatedwithCSFAβ.

Unfortunately, our studies corroborate other reports that BACE1 substrate NTFs may bind

plaques inthebrain (Bayer,Cappaietal.1999), thuscompromisingtheirbiomarkerpotential

overthecourseofAD.Itispossible,however,thattheycouldprovideutilityinearly(pre-plaque)

ADpatientsorinplaque-freeneurodegenerativedisorderssuchasParkinson’sdisease(PD).

4.3.2.2TotalTauasaBiomarkerofAD-RelatedEndosomalDysfunction

Although tau is known to be secreted from the endolysosomal pathway, we consider the

observationthatretromer-dependentendosomaldysfunctioncausesanincreaseinCSFtauthe

more importantandnovelresult fromourstudies. Priorpatientstudiesprovidedsomeclues

aboutthisrelationship.CSFstudiesinADpatientscarryingdeleteriousgeneticvariantsencoding

theretromerreceptorSorl1haveshownthatthesevariantsareassociatedwithrelativeincreases

inCSFtau.However,nostudies,toourknowledge,haveprofiledtheCSFinSorl1carrierswho

are plaque free. Since amyloid plaques are avid binders of APP fragments andmany other

83

extracellular proteins, itwould be interesting to probe the CSF of carriers of Sorl1 causative

mutations, inpre-plaquestagesofdisease. Neimann-PicktypeC1(NPC1) isadisorderthat is

interesting in this regard, since it is adisorderwhoseabnormal cholesterol trafficking causes

primaryendosomaldysfunction,withoutcausingplaqueformation.Interestingly,thesepatients

havebeenfoundtohaveCSFelevationsinbothtauandAb. WhileNPC1patientsdodevelop

neurofibrillarytanglesitisunlikelythattheirintraneuronaltaupathologyaccountsfortheCSF

taufindings,sinceFTDpatientswithaggressivetaupathologydonottypicallyhaveincreasesin

CSFtau.

UnlikefragmentsofAPPandperhapsitshomologues,taudoesnotbindamyloidplaques,which

commonlyexist inevennormalaging individuals. Thus,amongtheproteinswe investigated,

elevation in CSF tau has the best potential of being a useful biomarker of AD-associated

endosomal dysfunction. Our studies show that CSF tau can act as a sensitive marker of

endosomaldysfunction,butsayslittleaboutitsspecificity.AsabiomarkerofAD,chronicCSF

elevationintotaltauandptaudoesnotonlyhavehighsensitivity,butwhenobservedchronically

itappearstobespecific.Otherneurodegenerativediseases,withjustasmuchneuronaldeath

orNFTburdenasAD,donottypicallymanifestelevationsinCSFtau(Zetterberg2017),andso

AD’suniquepathophysiologymustdrivethiselevationandnotcelldeathortaupathology.While

acute events, such as strokes or seizures, can cause an increase in CSF tau, this elevation is

transientandnotchronic.

84

Whataspect,then,ofAD’spathophysiologyisthedriverofchronicelevationsinCSFtau?Since

tauissecretedfromtheendolysosomalsystemweanticipatethatanydefectthatincreasestau

accumulation in this compartment will play a role. Our studies permit us to conclude that

retromer-dependentendosomaltraffickingregulatesCSFtau.Itislikelythatotherdefectsinthe

endolysosomalsystemwillhaveasimilareffect.Additionally,defectsintauclearance,whilestill

poorlyunderstood,mustalsobeconsidered. Withtheavailabilityofnewdetectiontoolsand

reagentsthatmeasuremouseCSFtau,itisnowpossibletoinvestigatemultiplemousemodelsin

order to map the full range of molecular defects that might regulate CSF tau. With the

developmentoffuturereagentstomeasurenotonlytotaltau,butlevelsofitsfragmentsand

phosphorylation state, the uniquemolecular profile of CSF tau linked to differentmolecular

defectswillalsobeinformative.

AseriesofrecentstudiesillustratehowthecurrentabilitytodetecttotaltauinmouseCSFcan

elegantlybeusedtobeginclarifyingthesequestions.Thesestudieshaveshownthatmicethat

expressmutationsinAPPandPS1developelevationsinCSFtauovertime(Maia,Kaeseretal.

2013,Schelle,Hasleretal.2017).Sincethesemicedonotdevelopovertneurodegenerationor

tanglepathology, theyprovidedirect evidence that cell deathor taupathologyper se is not

requiredforCSFtauelevations.WhilethesestudieshaveshownthatblockingAPPcleavagewith

aBACE1inhibitorreducestauelevation(Schelle,Hasleretal.2017),implicatingAPPfragments,

howthesefragmentsleadtoincreasedtausecretionremainsunknown.Interestingly,however,

independent studies have shown that mice bearing these mutations develop retromer-

dependentdefectsovertime(ChuandPratico2017),andmorebroadlythattheaccumulationof

85

intraneuronal APP fragments cause endosomal dysfunction (Jiang,Mullaney et al. 2010, Xu,

Weissmilleretal.2016).

As recently reviewed,endosomaldysfunction representsaunifiedcellbiologicaldefect inAD

(Small, Simoes-Spassovetal.2017),whichcanbecausedprimarilybymoleculardefects that

directly target the endosome, or secondarily via defects that elevate intraneuronal APP

fragments. Moreover, evidence suggests that whether directly or indirectly, endosomal

dysfunction can act a pathogenic hub, mediating common downstream abnormalities. Our

resultswithretromerknockoutmiceshow,forthefirsttime,thatmoleculesthatdirectlytarget

the endosome lead to established abnormalities in CSF tau. The prior studieswithAPP/PS1

mutant mice established that increasing APP fragments phenocopies the same abnormality.

Whether in the latter case this occurs through the effect on APP fragments on endosomal

dysfunction,remainstobetested.Forthereasonslistedabove,thisseemshighlyplausible.If

so, then elevation in CSF tauwould be considered a sensitive and specific biomarker of AD-

associatedendosomaldysfunction.

86

5. Chapter5:Conclusionsandfuturedirections

5.1 Overview

AgrowingnumberofADtherapeuticshavefailedinclinicaltrials,promptingthequestfornew

drugtargets.Asaresult,endosomaltraffickinghasemergedasanewlyvalidatedtargetfordrug

discovery.Currently lacking,however,arebiomarkersofendosomaldysfunction,essential for

addressingtheepidemiologyofthiscellulardeficitinADandfortrackingitscourseoverdisease

progressionandinclinicaltrials.

We sought to fulfill this urgent need using retromer deficiency as a model of AD-related

endosomaldysfunction.Accordingly,weperformedabatteryofbiochemicalanalysesonCSFand

brainexosomesofmicewithselectivedeletionofretromer’scoreproteins.Ourstudiesinclude

the first characterizationof themouseCSF lipidomeand thedeepest characterizationof the

mouseCSFproteometodate,andproposeaCSFproteinbiomarkerwithclinicalutility.

5.2 CSFlipidomics

VPS26ahaploinsufficiencyresultsinamildCSFsphingolipidprofileofnon-neuronalorigin,while

completeknockoutofVPS26ainforebrainneuronsyieldsnolipidomicprofileatall.Whetherthis

indicatesthattheVPS26a-containingassemblyismorecriticaltonon-neuronalcellsorthatnon-

neuronalcellscontributemoretothemurineCSFlipidomecannotbeconcludedfromthisstudy.

Yet,itwarrantsconsiderationthattranslationofthisprofiletohumanCSFislimitedentirelyto

retromerdysfunctioninnon-neuronalcells(ofayetundeterminedtype).Furtherclarificationof

suchabiomarkerrequiresdissectionoftheextracellularlipidomiccontributionbyvariousglial

87

types.Thiscouldpotentiallybeexploredinvitrothroughlentiviralknockdownofretromerincell

type-specificcultures(e.g.astrocytes,microglia).Alternatively,invivostudiesofVPS26a-floxed

miceusingCre-recombinaseexpressionunderothercelltype-specificpromoters(e.g.CX3CR1for

microglia)arewarranted.AssphingomyelinstudiesinhumanCSFtodatehaveyieldedconflicting

results and our CSF proteomic experiments unearthedmore compelling leads, this lipidomic

profileofnon-neuronalretromerdysfunctionwasnotpursuedfurther.

Possiblemechanisms for these changesmight includemistraffickingofbiosyntheticenzymes.

Neutralsphingomyelinase,whichconvertssphingomyelinintoceramide,alsoregulatesexosomal

budding (Trajkovic,Hsuet al. 2008).Asourbrain-derivedexosomal studies suggest retromer

mightimpactexosomalbiogenesis,enhancedendosomalretentionofthissphingomyelinaseis

anintriguinghypothesis.

5.3 Brainexosomestudies

Exploration of two retromer knockoutmodels (full body knockout of VPS26b and forebrain-

selective neuronal knockout of VPS35) revealed that the amyloidogenic effects of retromer

knockoutinneurons(viaincreasedretentionofbothBACE1andAPPintheendosome)canbe

detectedinbrainexosomes.Thisenrichmentofb-CTFsinexosomes,whichtakeoriginfromthe

endosomalmembrane,isperhapsunsurprisinggiventheabove-mentionedeffectsofretromer

dysfunctiononamyloidogenicprocessingpathways.

88

Follow-upstudiesinhumanserumexosomesofbrainorigin,however,failedtodetectb-CTFsin

thesevesicles.Whether(a)thesefragmentsdonotendupinserum-circulatingexosomesor(b)

they simply cannot be detected using current technology is unknown. Studies to clarify this

questionarewarrantedinthepursuitofabrainexosomebiomarker.Theformerwouldimpose

adifferentstrategy,perhapsusingserumexosomesinneuronal-selectiveretromerknockoutsas

a starting material for biomarker discovery to overcome the challenges associated with

circulation entry. The latterwould prompt development ofmore sensitive technologies (e.g.

simoa)fordetectionofbrain-derivedproteinsinserumexosomes.

Anunexpectedfindingduringthecourseofthesestudieswasthepotentialimpactofretromer

deficiencyonexosomebiogenesis.Thiswassuggestedbyreductionofexosomalmarkersalixand

flotillin-1andapossible(asyetdescriptive)shifttowardsmallervesicles. Interestingly,recent

workbyanothergroupindicatesthatbrain-derivedexosomesfromADpatientsaresmallerand

bearfewercanonicalexosomalmarkersthanthoseofmatchedcontrols(Tran,Mustapicetal.).

While theseparallels are intriguing, further clarificationof retromer’s role in theseexosomal

biogenic pathways is required. Future studies would likely include electron microscopy and

nanotracker analysis to precisely assess changes in morphology of both exosomes and the

originatingMVBs.Inaddition,proteomicstudiescoulduncoverchangesinexosomalcontentand

perhapsprovidemechanisticclues.Thoughourbiomarkerstudiesbecamefocusedonthemore

compellingCSFproteomicfindings,theimpactofretromerdeficiencyonexosomalbiogenesisis

acontinuedinterestofthelab,largelymotivatedbytheresultsofthisproject.

89

5.4 CSFproteomics

ItisinterestingtonotethathaploinsufficiencyofVPS26ayieldsnodetectableproteinchangesin

theCSFatthreemonths.Perhaps,as inlipidomics,a latertimepointwouldintensifychanges

whichmayhavebeenmasked in this screen.Orperhapsa completeknockout is required to

detectproteomicchangesintheCSFofthesemice.Consideringthesepossibillities,weoptedto

makebothoftheseadjustmentsthroughanalysisofacompleteknockoutofVPS35inforebrain

neuronsatsixmonthsofage,summarizedbelow.

KnockoutofVPS35inforebrainneuronsresultsinasignificantchangeinnumerousCSFproteins,

includingtwocategoriesparticularlyrelevanttoADpathology:BACE1substrate(APLP1,APLP2,

CHL1, and APP) NTFs and Mapt (tau). Follow-up studies in vivo and in vitro confirmed the

elevation of BACE1 substrate NTFs and total tau in retromer knockout CSF. Interestingly,

IngenuityPathwayAnalysispositionedthesetwocategories(APPandtau)asthetopupstream

‘regulators’ofallCSFchanges,suggestingtheirrelevanceinthedownstreameffectsofretromer-

dependent endosomal dysfunction. Follow-up studies in CSF from both mice and humans

revealed striking correlations between levels of BACE1-cleaved fragments, implicating BACE1

modulation(e.g.byretromerdysfunction)asaprimarydriveroftheirextracellularrelease,which

canbedirectlymeasuredinCSF.

Perhapsmostremarkable,however,wasthecorrelationbetweentheNTFofBACE1substrate

APLP1 and total tau in plaque-free human CSF, suggesting biological convergence (at the

intracellularendosome)oftheseseeminglydistinctcategoriesofhits.Yet,follow-upstudiesina

90

smallcohortofplaque-positiveADpatientssupportearlierreportsthattheseAPP-relatedBACE1

substrateNTFsmaybindextracellularamyloid.Andalthoughthesefragmentsmightstillcarry

biomarker potential in early (pre-plaque) patients or in other plaque-free neurodegenerative

diseases such as PD, thier utility is therefore compromised in the progressive AD brain.

Accordingly, CSF total tau emerges as the top validated hit as a biomarker of AD-related

endosomaldysfunctionwithutilitythroughoutthecourseofdiseaseprogression.Remarkably,

studies from other groups have shown that APPmice, which undergo progressive retromer

deficiency,alsoexhibitage-relatedelevationsinCSFtotaltau(Maia,Kaeseretal.2013).

Nonetheless, follow-up studies are required to confirm the specificity of CSF total tau as a

biomarkerofendosomaldysfunctionandthatagentswhichrescuethiscellulardeficitreduceCSF

total tau. To address this question, investigationsofothermodelsofAD-relevant endosomal

dysfunction are required. Various models might include neuronal-selective knockouts of

Phosphatidylinositol3-kinaseVPS34(Miranda,Lasieckaetal.2018)andmicebearingtheAPOE4

allele (Nuriel, Peng et al. 2017). Additionally, treatmentwith retromer-stabilizing chaperones

(Berman,Ringeetal.2015)orviraldeliveryofretromercoreproteinswillaidthismission.

Post-translational modifications of tau are numerous and include various cleavage and

phosphorylationevents.WorkinbothmiceandhumanssuggeststhatthemajorityofCSFtauis

at least partially truncated at the C-terminal (and perhaps the N-terminal, as well) (Barten,

Cadelinaetal.2011,Meredith,Sankaranarayananetal.2013).Thetoolsweusedinthisstudy

rely upon detection of a central epitope, as do standard assessments of CSF total tau in AD

91

patients. Future work to clarify the exact sequencemay impart additional specificity to this

markerwhilealso informingon theunderlyingbiology.Mass spectrometryofneuronculture

mediawillgreatlyassistthisinquiry(asmurineCSFtauisnearlyundetectable,eveninretromer

knockoutmice).

WhileatleastaportionofCSFtauisphosphorylated,thestrikingcorrelationsbetweenAPLP1

NTFsandtotaltauwenotedinhumanCSFwerelesssignificantinptau.Therefore,wepostulate

that thepathways involved in the secretionofAPLP1NTFsandphosphorylated taumightbe

more distinct andwe predict that the elevation in extracellular tau we see in our retromer

knockoutisnotdrivenentirelybyanincreaseinptau181.Follow-upworkinneuronalcultureis

requiredtoconfirmwhetherthisisindeedtrue.

Although the general objective of this project was limited to biomarker identification, the

identificationofsuchanintriguingandAD-relevantbiomarkermotivatesanincreasinginterest

in theunderlyingbiology.Theendolysosomal systemhasbeenclearly linked to tausecretion

(Saman,Kimetal.2012,Rodriguez,Mohamedetal.2017,Caballero,Wangetal.2018).Yet,it

remainsunclearwhichofthevariouscompartmentscontributemosttothisprocess.Recentwork

hasdemonstrated that tau canaccumulate in lateendosomes (Caballero,Wanget al. 2018),

compartmentswhichcanfusewiththeplasmamembrane,liberatingtheirlumenalcontentsinto

theextracellularspace.Othergroupshavedemonstratedthattaucanaccumulateinexosomes

(Wang,Balajietal.2017);yet,thisvesicle-boundpopulationcomprisesaverysmallpercentage

(2%)ofextracellulartau(Wang,Balajietal.2017).Thus,weproposethatthemorethan5-fold

92

increaseinextracellulartauinretromerknockoutCSFislikelynon-exosomal.Nonetheless,we

hypothesizethatitistheselateendosomes/MVBswhichgiverisetothetauaccumulationwesee

inbothCSFandculturemedia.Furtherinvestigationstopinpointthismechanismarewarranted

andincludebothimmunocytochemistryandelectronmicroscopy.

Thiscomprehensivebiomarkerdiscoveryprojectmadeuseofanassortmentofmousemodelsof

AD-relevantendosomaldysfunctiontouncovervariousbiomarkercandidatesinbrainexosomes

andCSF.Translationstohumans,however, isnotalwaysstraight-forward,despitesubstantial

genomic homology. The hurdle is further challenged here by the presence of extracellular

plaques which are known to alter the CSF proteome of AD patients. Our translational

explorationsofvariouscandidatessuggestthatCSFtotaltau—unaffectedbythepresenceofAD

plaques—is themost reliable biomarker of retromer-dependent endosomal dysfunctionwith

clinicalutility.Furthermore,iffollow-upstudiesconfirmsensitivityandspecificitythismolecular

profile,thenperhapsthemostremarkableconclusionofthisstudyisthatendosomaldysfunction

isauniversalfeatureofADpathology,presentingattheearliestofstages.

93

6. MaterialsandMethods

AntibodiesandReagents

Thefollowingantibodieswereusedforimmunoblot:VPS35(ab10099,mouse,1:1000,Abcam),

VPS26a (ab23892, rabbit, 1:1000, Abcam), VPS26b (NBP1-92575, rabbit, 1:500, Novus

Biologicals),VPS29(SAB2501105,goat,1:500,SigmaAldrich),Actin(NB600-535,mouse,1:10000,

NovusBiologicals),Albumin(NB600-41532,goat,1:5000,NovusBiologicals),CHL1(AF2147-SP,

goat,1:500,NovusBiologicals),CHL1(25250-1-AP,rabbit,1:1000,ProteinTech),APLP1(AF3179-

SP, rabbit, 1:500, R&D Systems), Tau (ab80579, mouse, 1:500, Abcam), beta-III tubulin

(ab107216,chicken,1:5000,Abcam),anti-mouseHRP(170-6516,Bio-Rad),anti-goatHRP(172-

1034, rabbit, 1:3000,Bio-Rad), anti-rabbitHRP (1706515, goat, Bio-Rad), anti-chicken800CW

(925-32218, donkey, 1:20000, Licor), anti-mouse 680LT (925-68022, donkey, 1:20000, Licor),

anti-rabbit(925-32211,goat,1:20000,Licor),ProteinG-HRP(18-161,1:10000,SigmaAldrich),

ProteinA-HRP(101023,1:10000,ThermoFisher).

ThefollowingELISAkitswereusedformouseandhumanCSFexperiments:HemoglobinMouse

ELISA Kit (ab157715, Abcam), human APLP1 (LS-F8957-1, LS Bio). Roche lumi-light substrate

(12015200001,SigmaAldrich)wasusedforchemiluminescence.

Mousemodels

AllmicewerehousedinabarrierfacilityandusedinaccordancewithNationalInstituteofHealth

andColumbiaUniversityInstitutionalAnimalCareandUseCommitteeregulations.

94

VPS35fl/fl;Cam2a-Creknockoutmice

MiceexpressingloxP-flankedVPS35(VPS35fl/fl)onaC57BL/6backgroundwerecrossedwithmice

expressing Cre-recombinase under the Camk2a promoter to knock-out VPS35 in forebrain

neurons.VPS35fl/flmicelackingCre-recombinaseservedasthecontrolcondition.Miceages6-12

monthswereusedforallexperiments.

VPS26afl/fl;Cam2a-Creknockoutmice

MiceexpressingloxP-flankedVPS26a(VPS26afl/fl)onaC57BL/6backgroundwerecrossedwith

miceexpressingCre-recombinaseundertheCamk2apromotertoknock-outVPS26ainforebrain

neurons.VPS26afl/flmicelackingCre-recombinaseservedasthecontrolcondition.Six-month-old

micewereusedforalllipidomicsexperiments.

VPS26b-/-Mice

VPS26b heterozygote were generated by a targeting vector containing a neomycin-resistant

expressioncassetteflankedbygenomicregionsofVPS26b,negativeselection,electroporation

intoembryonicstemcells,andinjectionintomouseblastocyst(asdescribedpreviously)(Kim,Lee

etal.2010).ChimericmaleswerematedwithC57BL/6femalemicetogenerateaheterozygote

colony(Kim,Leeetal.2010).Heterozygoteswerethenmatedtogeneratethecompleteknockout

andthewildtypecontrols.Twelve-month-oldmicewereusedforallexosomeexperiments.

VPS26a+/-Mice

CongenicVPS26aheterozygoteswerecrossedona120/SvEvagoutibackground

95

for10generationsandmaintainedthroughsiblingmating(Muhammad,Floresetal.2008).Mice

ages3-6monthswereusedforalllipidomicsandproteomicsexperiments.

MurineCSFcollection

MurineCSFwascollected inapost-mortemprocedure followingCO2overdose inaccordance

with IACUCguidelines.Briefly,euthanizedmicewereplaced inapronepositionand the skin

coveringthebackoftheneckwassurgicallyexcised.Acottonswabcontaining70%ethanolwas

usedtoremoveanyhairorskin fromtheexposedtissue.Then,a27-gaugeneedle (SV*27EL,

TerumoMedicalProducts)attachedtoa1-mlsyringe(329650,BDBiosciences)wasinsertedinto

thecisternamagnaallowingflowofCSFintothebutterflyneedle.After10-15seconds,theneedle

wasremovedandtheCSFaspiratedintomicrocentrifugetubes(1605-0000,USAScientific)which

wereimmediatelyplacedondryiceandsubsequentlystoredat-80C.Roughly5-10uLoffluidwas

collectedpermouse.CSFwhichwasvisiblycontaminatedwithbloodwasdiscarded.Allremaining

samplesunderwentmorestringentassessmentforbloodcontaminationviahemoglobinELISA

using0.5uL ina1:200dilution. Samplesbelow0.01%bloodcontaminationwere retained for

biochemicalanalyses.Foreachassay,bothindividualandpooledreplicatesbetweenconditions

were matched for hemoglobin content to reduce the likelihood of a blood contamination

confound.

Lipidomicanalyses

Lipidswereextracted fromequalvolumesofmurineCSF (30ul forpooledsamplesor7uL for

individual samples). Lipid extractswere prepared via chloroform-methanol extraction, spiked

96

withappropriateinternalstandards,andanalyzedusinga6490TripleQuadrupoleLC/MSsystem

(Agilent Technologies, Santa Clara, CA) as described previously(Chan, Oliveira et al. 2012).

Glycerophospholipids and sphingolipids were separated with normal-phase HPLC using an

Agilent Zorbax Rx-Sil column (inner diameter 2.1 x 100mm) under the following conditions:

mobilephaseA(chloroform:methanol:1Mammoniumhydroxide,89.9:10:0.1,v/v/v)andmobile

phaseB(chloroform:methanol:water:ammoniumhydroxide,55:39.9:5:0.1,v/v/v);95%Afor2

min,lineargradientto30%Aover18minandheldfor3min,andlineargradientto95%Aover

2minandheldfor6min.Quantificationoflipidspecieswasaccomplishedusingmultiplereaction

monitoring(MRM)transitionsthatweredevelopedinearlierstudies(Chan,Oliveiraetal.2012)

in conjunction with referencing of appropriate internal standards: ceramide d18:1/17:0 and

sphingomyelind18:1/12:0(AvantiPolarLipids,Alabaster,AL).Valuesarerepresentedasmole

fractionwithrespecttototallipid(%molarity).Forthis,lipidmass(inmoles)ofanyspecificlipid

isnormalizedbythetotalmass(inmoles)ofallthelipidsmeasured(Chan,Oliveiraetal.2012).In

addition,allofourresultswerefurthernormalizedbyproteincontent.

Proteomicanalyses:LabeledLC-MS/MSofVPS26a+/-CSF

HighqualityCSF(<0.01%bloodcontamination)collectedfromVPS26+/-andwildtypelittermates

was pooled into four biological replicates per genotype of 30uL each. Each replicate was a

composite of CSF from roughly 4-6 mice. Protein concentrations from each sample were

determined by a fluorescent-based protein assay (Q33211, Life Technology). Concentrations

rangedfrom0.68-0.8µg/mLpersample.Equalamounts(20μg)fromeachsampleweredigested

by trypsinand labeledwithTMT isobaricmass tags.A reference sample (comprisedofequal

97

amounts fromeachsample)wasgeneratedforquantification.AfterTMT labeling,allsamples

were mixed and peptides were fractionated by C18 reverse phase with a pH gradient (Life

Technologies).PeptidemixtureswereanalyzedbyOrbitrapFusionTribridmass spectrometer

(Thermo Fisher) and searched against the mouse Uniprot protein database using Proteome

Discoverer2.1 (ThermoFisher).Overall intensityofeachTMTchannelwasnormalized to the

referencesample.Minitab17wasusedtoperformstatisticalanalysesandgeneratestatistical

plots. Qlucore Omic Exporer software was used to perform statistical analysis on

phosphopeptides.

Proteomicanalyses:UnlabeledLC-MS/MSofVPS35Camk2aKnockoutCSF

High quality CSF (<0.01% blood contamination) collected from VPS35 knockout and control

littermateswaspooledintobiologicalreplicatesof30uLeach,whichwereroughlymatchedfor

hemoglobin content. Each replicate was a composite of CSF from 4-7 mice. The protein

concentrationofeachsamplewasdeterminedviaQubitProteinQuantificationAssay(Q33211,

LifeTechnology).Concentrationsforeachsamplerangedfrom0.10-0.16mg/ml.Eachsamplewas

loadedontoanS-trapcolumn,digestedwithtrypsin,elutedwithtrifluoroaceticacid,anddried.

Each sample was divided into two technical replicates (each containing 1.5ug of digested

protein).MS/MSwasperformedoneachsampleusingaThermoOrbitrapFusionTribridMass

Sepctrometer(ThermoFisher).ProteomeDiscoverersoftware(version1.4)wasusedtosearch

theacquitedMS/MSdataagainstamouseproteindatabasedownloadedfromUniProt.Positive

identificationwassetat1%proteinFDRforthoseproteinshavingatleasttwoquantifiablevalues

detected.Anyduplicatedprotein identifications from thedatabasewere removed.A totalof

98

1505 mouse proteins were detected and included in the final data set. For each protein,

quantification was determined by averaging the peak area of the threemost abundant and

distinctpeptides.Totalareaswereusedfornormalization.QlucoreOmicsExplorerandMinitab17

Softwarewereused toperformcorrelationand statistical analyses. Spearmancorrelationsof

technical replicates were used to determine group outliers. One outlier per genotype was

identifiedbytheProteomicsCoreandremovedpriortodownstreamanalyses.

GenerationofGFP.CreandGFP.D-CreVirusforVPS35KnockoutinCellCulture

HEK293Tcellsweregrownto70%confluence inmediacontainingDMEM,5%glutaMAx,10%

FBS,andpen/streppriortolipofectamine(LTX)transfectionwithplasmidscontainingD8.2,VSVG,

andeitherGFP.CreorGFP.D-Cre.After72hours,mediawascollectedandfiltered(SE1M003M00,

Millipore)beforeincubationwithlenti-xconcentrator(631231,Clontech)andcentrifugationto

collecttheviralparticles.

VPS35KnockoutinPrimaryNeuronCultures

HippocampiorcerebralcorticesfromVPS35fl/flP0pupsweredissectedincoldHBSSanddigested

in 0.5% Trypsin-EDTA (25200056, Invitrogen) for 20minutes at 37C. The digested tissuewas

rinsed in HBSS and homogenized gently with a flame-polished glass pipette in NBA media

containing Glutamax (35050-061, Life Technologies), B-27 (17504044, Thermo Fisher), and

pen/strep(15140-122,ThermoFisher).Neuronswereplatedinsix-wellplates(83.3921,Sarstedt)

containing poly-D-lysine-coated 12mm glass coverslips (NC9702891, Thermo Fisher) for

immunofluorescence or in poly-L-ornithine-coated 12-well plates (83.3921, Sarstedt) for

99

immunoblotandmediacollectionexperiments.Mediawasreplacedat24hoursand1/3media

replacements were conducted every 5 days thereafter. On DIV5, lentivirus containing either

GFP.CreorGFP.D-Creviruswasaddedtoeachwell.OnDIV15,mediawasremovedandneurons

fixed in 4% paraformaldehyde (15710-S, Electron Microscopy Science) for downstream

immunofluorescenceorlysedincoldRIPAbuffer(89901,ThermoFisher)forimmunoblot.

Immunoblotting

SampleswerepreparedinLDSsamplebuffer(NP0007,ThermoFisher)andreducingagent(161-

0792,Bio-Rad),boiled for5minutesat90Cand runviaelectrophoresis in4-12%bis-tris gels

(NP0336orNP0335, Thermo Fisher) inMESbuffer (NP0002-02, Thermo Fisher). For CSF and

mediaexperiments,equalvolumes(5-15uL)wereloadedineachlane;forlysate,equalmasses

wereloadedineachlane.wereloadedperlane.Proteinswerethentransferredtonitrocellulose

membranes(IB3010-31,Invitrogen)viaiBlot(P3;7:40)andblockedfor1hourinbovineserum

albumin (A30075-100gm, Research Product International) or odyssey blocking buffer (927-

40000, Licor) diluted 1:1 in PBS-T. Membranes were then probed with primary antibodies

overnight at 4C, washed in PBS-T, probedwith secondary antibodies (HRP-conjugated or IR-

conjugated) foronehouratroomtemperature.Membraneswereagainwashed inPBS-Tand

imagedinRochesubstrate(12015196001,SigmaAldrich)forECLoralonewiththeLicorodyssey

system.

TotalTauQuantificationinCultureMedia

100

Total tauwasquantified fromculturemediausing theMSDmouse total tau kit (K151DSD-1,

MesoScaleDiscovery)accordingtothemanufacturer’sguidelines.Wellswerefirstblockedfor

onehourandwashed.Then,25uLofculturemediawasaddedinduplicatewellsalongwiththe

total taustandardsprepared inNBAmedia(withglutamax,B-27,andpen-strep,asdescribed

above),andincubatedforonehour.Wellswerewashedandincubatedwithdetectionantibody

for an additional hour before washing and analyzing in read buffer T solution on the MSD

instrument.CelldeathwasassessedviaLDHlevelsintheculturedmediausingCytoTox-ONE™

HomogeneousMembraneIntegrityAssay(G7890,Promega).ValuesfortotaltauandLDHwere

normalizedtototalprotein,whichwasmeasuredusingBCA(23227,Pierce).

TotalTauQuantificationinMurineCSF

CSF total tauwasquantifiedby SIMOAusing themouse total tauassay (102209,Quanterix).

Individual CSF samples (5uL)werediluted1:80 in samplebuffer and then split into technical

duplicates.Standardsandsampleswererunaccordingtothemanufacturer’s instructionsand

readontheHD-1analyzer(Quanterix).

HumanCSFStudies

CSFfrompatientswithAlzheimer’sdiseaseorMildCognitiveImpairment,andNormalControls,

wasrequestedfromtheColumbiaUniversityAlzheimer’sDiseaseResearchCenter(ADRC)CSF

bank. This CSF bank includes CSF obtained by lumbar puncture and banked for research

accordingtoprotocolsapprovedbytheColumbiaUniversityIRB.CSFwasstoredfrozenin400uL

aliquots,inpolypropylenetubes,at-80C.FrozenCSFaliquotswerelaterthawed,andanalyzed

101

forAb42,totaltau,andphospho-tau,bytheADRCClinicalCore,usingamultiplexfluorimetric

microspherexMAPbead-basedsandwichimmunoassayusingmonoclonalantibodiescovalently

coupledtospectrallyspecificfluorescentbeads.Thisassay(Inno-BIAAlzBio3,FujirebioUSInc,

Alpharetta,GA)wasperformedusingaLuminex100machine.IdenticalfrozenCSFaliquotswere

thawedandusedfortheexperimentsdescribedhere.N-APLP1andN-CHL1weremeasuredvia

immunoblotasdescribedabove.N-APLP1werealsomeasuredviaELISA(LS-F8957-1,LSBio)in

duplicateaccordingtothemanufacturer’sinstructionsfollowinga1:20dilutionasinformedby

early optimizations in control CSF. Total protein was calculated via BCA (23227, Pierce).

Thresholds implementedtoconfirmplaquepositivity inADpatients includesAb42<210pg/mL

andtotaltau/Ab42>0.39.

Statisticalmethods

Student’s t-testwas used for all biochemical experiments and immunohistochemical analysis

usingtwo-taileddistributionwithequalvariance(P<0.05).Limmasoftwarewasusedforanalysis

ofCSFproteomics,andFDR-correctedq-valuesare reported.Significance is indicatedas*P<

0.05,**P<0.01and***P<0.001;****P<0.0001).

102

7. References

Abner,E.L.,G.A.Jicha,L.M.Shaw,J.Q.TrojanowskiandE.J.Goetzl(2016)."PlasmaneuronalexosomallevelsofAlzheimer'sdiseasebiomarkersinnormalaging."AnnClinTranslNeurol3(5):399-403.

Alonso,A.,T.Zaidi,M.Novak,I.Grundke-IqbalandK.Iqbal(2001)."Hyperphosphorylationinducesself-assemblyoftauintotanglesofpairedhelicalfilaments/straightfilaments."ProcNatlAcadSciUSA98(12):6923-6928.

Altmann,A.,B.Ng,S.M.Landau,W.J.Jagust,M.D.GreiciusandI.Alzheimer'sDiseaseNeuroimaging(2015)."Regionalbrainhypometabolismisunrelatedtoregionalamyloidplaqueburden."Brain138(Pt12):3734-3746.

Altmann,A.,L.Tian,V.W.Henderson,M.D.GreiciusandI.Alzheimer'sDiseaseNeuroimagingInitiative(2014)."SexmodifiestheAPOE-relatedriskofdevelopingAlzheimerdisease."AnnNeurol75(4):563-573.

Alvarez-Erviti,L.,Y.Seow,A.H.Schapira,C.Gardiner,I.L.Sargent,M.J.WoodandJ.M.Cooper(2011)."Lysosomaldysfunctionincreasesexosome-mediatedalpha-synucleinreleaseandtransmission."NeurobiolDis42(3):360-367.

Andreasen,N.,L.Minthon,P.Davidsson,E.Vanmechelen,H.Vanderstichele,B.WinbladandK.Blennow(2001)."EvaluationofCSF-tauandCSF-Abeta42asdiagnosticmarkersforAlzheimerdiseaseinclinicalpractice."ArchNeurol58(3):373-379.

APA.(2017)."WhatisAlzheimer’sDisease?",fromhttps://www.psychiatry.org/patients-families/alzheimers/what-is-alzheimers-disease.

Archer,H.A.,P.Edison,D.J.Brooks,J.Barnes,C.Frost,T.Yeatman,N.C.FoxandM.N.Rossor(2006)."AmyloidloadandcerebralatrophyinAlzheimer'sdisease:an11C-PIBpositronemissiontomographystudy."AnnNeurol60(1):145-147.

Arighi,C.N.,L.M.Hartnell,R.C.Aguilar,C.R.HaftandJ.S.Bonifacino(2004)."Roleofthemammalianretromerinsortingofthecation-independentmannose6-phosphatereceptor."JCellBiol165(1):123-133.

Armstrong,A.,N.Mattsson,H.Appelqvist,C.Janefjord,L.Sandin,L.Agholme,B.Olsson,S.Svensson,K.Blennow,H.ZetterbergandK.Kagedal(2014)."LysosomalnetworkproteinsaspotentialnovelCSFbiomarkersforAlzheimer'sdisease."NeuromolecularMed16(1):150-160.

Asai,H.,S.Ikezu,S.Tsunoda,M.Medalla,J.Luebke,T.Haydar,B.Wolozin,O.Butovsky,S.KuglerandT.Ikezu(2015)."Depletionofmicrogliaandinhibitionofexosomesynthesishalttaupropagation."NatNeurosci18(11):1584-1593.

103

Ashford,J.W.(2004)."APOEgenotypeeffectsonAlzheimer'sdiseaseonsetandepidemiology."JMolNeurosci23(3):157-165.

Askanas,V.,W.K.Engel,M.Bilak,R.B.AlvarezandD.J.Selkoe(1994)."TwistedtubulofilamentsofinclusionbodymyositismuscleresemblepairedhelicalfilamentsofAlzheimerbrainandcontainhyperphosphorylatedtau."AmJPathol144(1):177-187.

Attar,N.andP.J.Cullen(2010)."Theretromercomplex."AdvEnzymeRegul50(1):216-236.

Bales,K.R.,T.Verina,D.J.Cummins,Y.Du,R.C.Dodel,J.Saura,C.E.Fishman,C.A.DeLong,P.Piccardo,V.Petegnief,B.GhettiandS.M.Paul(1999)."ApolipoproteinEisessentialforamyloiddepositionintheAPP(V717F)transgenicmousemodelofAlzheimer'sdisease."ProcNatlAcadSciUSA96(26):15233-15238.

Barten,D.M.,G.W.Cadelina,N.Hoque,L.B.DeCarr,V.L.Guss,L.Yang,S.Sankaranarayanan,P.D.Wes,M.E.Flynn,J.E.Meredith,M.K.AhlijanianandC.F.Albright(2011)."Tautransgenicmiceasmodelsforcerebrospinalfluidtaubiomarkers."JAlzheimersDis24Suppl2:127-141.

Bayer,T.A.,R.Cappai,C.L.Masters,K.BeyreutherandG.Multhaup(1999)."Itallstickstogether--theAPP-relatedfamilyofproteinsandAlzheimer'sdisease."MolPsychiatry4(6):524-528.

Bechler,M.E.,A.M.Doody,K.D.Ha,B.L.Judson,I.ChenandW.J.Brown(2011)."ThephospholipaseA(2)enzymecomplexPAFAHIbmediatesendosomalmembranetubuleformationandtrafficking."MolBiolCell22(13):2348-2359.

Bejanin,A.,D.R.Schonhaut,R.LaJoie,J.H.Kramer,S.L.Baker,N.Sosa,N.Ayakta,A.Cantwell,M.Janabi,M.Lauriola,J.P.O'Neil,M.L.Gorno-Tempini,Z.A.Miller,H.J.Rosen,B.L.Miller,W.J.JagustandG.D.Rabinovici(2017)."TaupathologyandneurodegenerationcontributetocognitiveimpairmentinAlzheimer'sdisease."Brain140(12):3286-3300.

Berman,D.E.,D.Ringe,G.A.PetskoandS.A.Small(2015)."TheuseofpharmacologicalretromerchaperonesinAlzheimer'sdiseaseandotherendosomal-relateddisorders."Neurotherapeutics12(1):12-18.

Bhalla,A.,C.P.Vetanovetz,E.Morel,Z.Chamoun,G.DiPaoloandS.A.Small(2012)."Thelocationandtraffickingroutesoftheneuronalretromeranditsroleinamyloidprecursorproteintransport."NeurobiolDis47(1):126-134.

Bird,T.D.(1993).Early-OnsetFamilialAlzheimerDisease.GeneReviews((R)).M.P.Adam,H.H.Ardinger,R.A.Pagonetal.Seattle(WA).

Bitan,G.,M.D.Kirkitadze,A.Lomakin,S.S.Vollers,G.B.BenedekandD.B.Teplow(2003)."Amyloidbeta-protein(Abeta)assembly:Abeta40andAbeta42oligomerizethroughdistinctpathways."ProcNatlAcadSciUSA100(1):330-335.

104

Bloom,G.S.(2014)."Amyloid-betaandtau:thetriggerandbulletinAlzheimerdiseasepathogenesis."JAMANeurol71(4):505-508.

Braak,H.andE.Braak(1991)."NeuropathologicalstageingofAlzheimer-relatedchanges."ActaNeuropathol82(4):239-259.

Bugarcic,A.,I.Vetter,S.Chalmers,G.Kinna,B.M.CollinsandR.D.Teasdale(2015)."Vps26B-retromernegativelyregulatesplasmamembraneresensitizationofPAR-2."CellBiolInt39(11):1299-1306.

Caballero,B.,Y.Wang,A.Diaz,I.Tasset,Y.R.Juste,B.Stiller,E.M.Mandelkow,E.MandelkowandA.M.Cuervo(2018)."Interplayofpathogenicformsofhumantauwithdifferentautophagicpathways."AgingCell17(1).

Capodivento,G.,D.Visigalli,M.Garnero,R.Fancellu,M.D.Ferrara,A.Basit,Z.Hamid,V.P.Pastore,S.Garibaldi,A.Armirotti,G.Mancardi,C.Serrati,E.Capello,A.SchenoneandL.Nobbio(2017)."SphingomyelinasamyelinbiomarkerinCSFofacquireddemyelinatingneuropathies."SciRep7(1):7831.

Carter,C.L.,E.M.Resnick,M.MallampalliandA.Kalbarczyk(2012)."SexandgenderdifferencesinAlzheimer'sdisease:recommendationsforfutureresearch."JWomensHealth(Larchmt)21(10):1018-1023.

Cataldo,A.M.,J.L.Barnett,C.PieroniandR.A.Nixon(1997)."IncreasedneuronalendocytosisandproteasedeliverytoearlyendosomesinsporadicAlzheimer'sdisease:neuropathologicevidenceforamechanismofincreasedbeta-amyloidogenesis."JNeurosci17(16):6142-6151.

Cataldo,A.M.,C.M.Peterhoff,J.C.Troncoso,T.Gomez-Isla,B.T.HymanandR.A.Nixon(2000)."EndocyticpathwayabnormalitiesprecedeamyloidbetadepositioninsporadicAlzheimer'sdiseaseandDownsyndrome:differentialeffectsofAPOEgenotypeandpresenilinmutations."AmJPathol157(1):277-286.

Chan,R.B.,T.G.Oliveira,E.P.Cortes,L.S.Honig,K.E.Duff,S.A.Small,M.R.Wenk,G.ShuiandG.DiPaolo(2012)."ComparativelipidomicanalysisofmouseandhumanbrainwithAlzheimerdisease."JBiolChem287(4):2678-2688.

Chartier-Harlin,M.C.,F.Crawford,H.Houlden,A.Warren,D.Hughes,L.Fidani,A.Goate,M.Rossor,P.Roques,J.Hardyandetal.(1991)."Early-onsetAlzheimer'sdiseasecausedbymutationsatcodon717ofthebeta-amyloidprecursorproteingene."Nature353(6347):844-846.

Chen,F.W.,R.E.GordonandY.A.Ioannou(2005)."NPC1lateendosomescontainelevatedlevelsofnon-esterified('free')fattyacidsandanabnormallyglycosylatedformoftheNPC2protein."BiochemJ390(Pt2):549-561.

105

Childs-Disney,J.L.,M.Wu,A.Pushechnikov,O.AminovaandM.D.Disney(2007)."AsmallmoleculemicroarrayplatformtoselectRNAinternalloop-ligandinteractions."ACSChemBiol2(11):745-754.

Choy,R.W.,Z.ChengandR.Schekman(2012)."Amyloidprecursorprotein(APP)trafficsfromthecellsurfaceviaendosomesforamyloidbeta(Abeta)productioninthetrans-Golginetwork."ProcNatlAcadSciUSA109(30):E2077-2082.

Choy,R.W.,M.Park,P.Temkin,B.E.Herring,A.Marley,R.A.NicollandM.vonZastrow(2014)."Retromermediatesadiscreterouteoflocalmembranedeliverytodendrites."Neuron82(1):55-62.

Chu,J.andD.Pratico(2017)."TheretromercomplexsysteminatransgenicmousemodelofAD:influenceofage."NeurobiolAging52:32-38.

Collino,F.,M.Pomatto,S.Bruno,R.S.Lindoso,M.Tapparo,W.Sicheng,P.QuesenberryandG.Camussi(2017)."ExosomeandMicrovesicle-EnrichedFractionsIsolatedfromMesenchymalStemCellsbyGradientSeparationShowedDifferentMolecularSignaturesandFunctionsonRenalTubularEpithelialCells."StemCellRev13(2):226-243.

Costa-Silva,B.,N.M.Aiello,A.J.Ocean,S.Singh,H.Zhang,B.K.Thakur,A.Becker,A.Hoshino,M.T.Mark,H.Molina,J.Xiang,T.Zhang,T.M.Theilen,G.Garcia-Santos,C.Williams,Y.Ararso,Y.Huang,G.Rodrigues,T.L.Shen,K.J.Labori,I.M.Lothe,E.H.Kure,J.Hernandez,A.Doussot,S.H.Ebbesen,P.M.Grandgenett,M.A.Hollingsworth,M.Jain,K.Mallya,S.K.Batra,W.R.Jarnagin,R.E.Schwartz,I.Matei,H.Peinado,B.Z.Stanger,J.BrombergandD.Lyden(2015)."Pancreaticcancerexosomesinitiatepre-metastaticnicheformationintheliver."NatCellBiol17(6):816-826.

Cozier,G.E.,J.Carlton,A.H.McGregor,P.A.Gleeson,R.D.Teasdale,H.MellorandP.J.Cullen(2002)."Thephoxhomology(PX)domain-dependent,3-phosphoinositide-mediatedassociationofsortingnexin-1withanearlysortingendosomalcompartmentisrequiredforitsabilitytoregulateepidermalgrowthfactorreceptordegradation."JBiolChem277(50):48730-48736.

Cutler,R.G.,J.Kelly,K.Storie,W.A.Pedersen,A.Tammara,K.Hatanpaa,J.C.TroncosoandM.P.Mattson(2004)."Involvementofoxidativestress-inducedabnormalitiesinceramideandcholesterolmetabolisminbrainagingandAlzheimer'sdisease."ProcNatlAcadSciUSA101(7):2070-2075.

DemoryBeckler,M.,J.N.Higginbotham,J.L.Franklin,A.J.Ham,P.J.Halvey,I.E.Imasuen,C.Whitwell,M.Li,D.C.LieblerandR.J.Coffey(2013)."ProteomicanalysisofexosomesfrommutantKRAScoloncancercellsidentifiesintercellulartransferofmutantKRAS."MolCellProteomics12(2):343-355.

Derivery,E.andA.Gautreau(2010)."EvolutionaryconservationoftheWASHcomplex,anactinpolymerizationmachineinvolvedinendosomalfission."CommunIntegrBiol3(3):227-230.

106

DiPaolo,G.andT.W.Kim(2011)."LinkinglipidstoAlzheimer'sdisease:cholesterolandbeyond."NatRevNeurosci12(5):284-296.

Dislich,B.,F.Wohlrab,T.Bachhuber,S.A.Muller,P.H.Kuhn,S.Hogl,M.Meyer-LuehmannandS.F.Lichtenthaler(2015)."Label-freeQuantitativeProteomicsofMouseCerebrospinalFluidDetectsbeta-SiteAPPCleavingEnzyme(BACE1)ProteaseSubstratesInVivo."MolCellProteomics14(10):2550-2563.

Duncan,H.D.,J.Nikelski,R.Pilon,J.Steffener,H.ChertkowandN.A.Phillips(2017)."StructuralbraindifferencesbetweenmonolingualandmultilingualpatientswithmildcognitiveimpairmentandAlzheimerdisease:Evidenceforcognitivereserve."Neuropsychologia109:270-282.

Edgar,A.J.andJ.M.Polak(2000)."Humanhomologuesofyeastvacuolarproteinsorting29and35."BiochemBiophysResCommun277(3):622-630.

Fagan,A.M.,M.A.Mintun,R.H.Mach,S.Y.Lee,C.S.Dence,A.R.Shah,G.N.LaRossa,M.L.Spinner,W.E.Klunk,C.A.Mathis,S.T.DeKosky,J.C.MorrisandD.M.Holtzman(2006)."InverserelationbetweeninvivoamyloidimagingloadandcerebrospinalfluidAbeta42inhumans."AnnNeurol59(3):512-519.

Fjorback,A.W.,M.Seaman,C.Gustafsen,A.Mehmedbasic,S.Gokool,C.Wu,D.Militz,V.Schmidt,P.Madsen,J.R.Nyengaard,T.E.Willnow,E.I.Christensen,W.B.Mobley,A.NykjaerandO.M.Andersen(2012)."RetromerbindstheFANSHYsortingmotifinSorLAtoregulateamyloidprecursorproteinsortingandprocessing."JNeurosci32(4):1467-1480.

Foley,P.(2010)."LipidsinAlzheimer'sdisease:Acentury-oldstory."BiochimBiophysActa1801(8):750-753.

Fonteh,A.N.,C.Ormseth,J.Chiang,M.Cipolla,X.ArakakiandM.G.Harrington(2015)."SphingolipidmetabolismcorrelateswithcerebrospinalfluidBetaamyloidlevelsinAlzheimer'sdisease."PLoSOne10(5):e0125597.

Ghossoub,R.,F.Lembo,A.Rubio,C.B.Gaillard,J.Bouchet,N.Vitale,J.Slavik,M.MachalaandP.Zimmermann(2014)."Syntenin-ALIXexosomebiogenesisandbuddingintomultivesicularbodiesarecontrolledbyARF6andPLD2."NatCommun5:3477.

Glabe,C.G.(2006)."Commonmechanismsofamyloidoligomerpathogenesisindegenerativedisease."NeurobiolAging27(4):570-575.

Glerup,S.,M.Lume,D.Olsen,J.R.Nyengaard,C.B.Vaegter,C.Gustafsen,E.I.Christensen,M.Kjolby,A.Hay-Schmidt,D.Bender,P.Madsen,M.Saarma,A.NykjaerandC.M.Petersen(2013)."SorLAcontrolsneurotrophicactivitybysortingofGDNFanditsreceptorsGFRalpha1andRET."CellRep3(1):186-199.

107

Goetzl,E.J.,A.Boxer,J.B.Schwartz,E.L.Abner,R.C.Petersen,B.L.Miller,O.D.Carlson,M.MustapicandD.Kapogiannis(2015)."LowneuralexosomallevelsofcellularsurvivalfactorsinAlzheimer'sdisease."AnnClinTranslNeurol2(7):769-773.

Goetzl,E.J.,A.Boxer,J.B.Schwartz,E.L.Abner,R.C.Petersen,B.L.MillerandD.Kapogiannis(2015)."Alteredlysosomalproteinsinneural-derivedplasmaexosomesinpreclinicalAlzheimerdisease."Neurology85(1):40-47.

Goetzl,E.J.,M.Mustapic,D.Kapogiannis,E.Eitan,I.V.Lobach,L.Goetzl,J.B.SchwartzandB.L.Miller(2016)."Cargoproteinsofplasmaastrocyte-derivedexosomesinAlzheimer'sdisease."FASEBJ30(11):3853-3859.

Gomez,T.S.andD.D.Billadeau(2009)."AFAM21-containingWASHcomplexregulatesretromer-dependentsorting."DevCell17(5):699-711.

Gong,C.X.andK.Iqbal(2008)."Hyperphosphorylationofmicrotubule-associatedproteintau:apromisingtherapeutictargetforAlzheimerdisease."CurrMedChem15(23):2321-2328.

Gonzales,P.A.,T.Pisitkun,J.D.Hoffert,D.Tchapyjnikov,R.A.Star,R.Kleta,N.S.WangandM.A.Knepper(2009)."Large-scaleproteomicsandphosphoproteomicsofurinaryexosomes."JAmSocNephrol20(2):363-379.

Gozal,Y.M.,D.Cheng,D.M.Duong,J.J.Lah,A.I.LeveyandJ.Peng(2006)."Mergeroflasercapturemicrodissectionandmassspectrometry:awindowintotheamyloidplaqueproteome."MethodsEnzymol412:77-93.

Grundke-Iqbal,I.,K.Iqbal,M.Quinlan,Y.C.Tung,M.S.ZaidiandH.M.Wisniewski(1986)."Microtubule-associatedproteintau.AcomponentofAlzheimerpairedhelicalfilaments."JBiolChem261(13):6084-6089.

Grundke-Iqbal,I.,K.Iqbal,Y.C.Tung,M.Quinlan,H.M.WisniewskiandL.I.Binder(1986)."Abnormalphosphorylationofthemicrotubule-associatedproteintau(tau)inAlzheimercytoskeletalpathology."ProcNatlAcadSciUSA83(13):4913-4917.

Guerreiro,R.,J.BrasandJ.Hardy(2013)."SnapShot:geneticsofAlzheimer'sdisease."Cell155(4):968-968e961.

Haft,C.R.,M.delaLuzSierra,R.Bafford,M.A.Lesniak,V.A.BarrandS.I.Taylor(2000)."HumanorthologsofyeastvacuolarproteinsortingproteinsVps26,29,and35:assemblyintomultimericcomplexes."MolBiolCell11(12):4105-4116.

Harbour,M.E.,S.Y.Breusegem,R.Antrobus,C.Freeman,E.ReidandM.N.Seaman(2010)."Thecargo-selectiveretromercomplexisarecruitinghubforproteincomplexesthatregulateendosomaltubuledynamics."JCellSci123(Pt21):3703-3717.

108

Hardy,J.andD.Allsop(1991)."AmyloiddepositionasthecentraleventintheaetiologyofAlzheimer'sdisease."TrendsPharmacolSci12(10):383-388.

Harrison,M.S.,C.S.Hung,T.T.Liu,R.Christiano,T.C.WaltherandC.G.Burd(2014)."Amechanismforretromerendosomalcoatcomplexassemblywithcargo."ProcNatlAcadSciUSA111(1):267-272.

Harterink,M.,F.Port,M.J.Lorenowicz,I.J.McGough,M.Silhankova,M.C.Betist,J.R.T.vanWeering,R.vanHeesbeen,T.C.Middelkoop,K.Basler,P.J.CullenandH.C.Korswagen(2011)."ASNX3-dependentretromerpathwaymediatesretrogradetransportoftheWntsortingreceptorWntlessandisrequiredforWntsecretion."NatCellBiol13(8):914-923.

Hebert,L.E.,J.Weuve,P.A.ScherrandD.A.Evans(2013)."AlzheimerdiseaseintheUnitedStates(2010-2050)estimatedusingthe2010census."Neurology80(19):1778-1783.

Holtzman,D.M.(2001)."Roleofapoe/AbetainteractionsinthepathogenesisofAlzheimer'sdiseaseandcerebralamyloidangiopathy."JMolNeurosci17(2):147-155.

Holtzman,D.M.,K.R.Bales,T.Tenkova,A.M.Fagan,M.Parsadanian,L.J.Sartorius,B.Mackey,J.Olney,D.McKeel,D.WozniakandS.M.Paul(2000)."ApolipoproteinEisoform-dependentamyloiddepositionandneuriticdegenerationinamousemodelofAlzheimer'sdisease."ProcNatlAcadSciUSA97(6):2892-2897.

Hoshino,A.,B.Costa-Silva,T.L.Shen,G.Rodrigues,A.Hashimoto,M.TesicMark,H.Molina,S.Kohsaka,A.DiGiannatale,S.Ceder,S.Singh,C.Williams,N.Soplop,K.Uryu,L.Pharmer,T.King,L.Bojmar,A.E.Davies,Y.Ararso,T.Zhang,H.Zhang,J.Hernandez,J.M.Weiss,V.D.Dumont-Cole,K.Kramer,L.H.Wexler,A.Narendran,G.K.Schwartz,J.H.Healey,P.Sandstrom,K.J.Labori,E.H.Kure,P.M.Grandgenett,M.A.Hollingsworth,M.deSousa,S.Kaur,M.Jain,K.Mallya,S.K.Batra,W.R.Jarnagin,M.S.Brady,O.Fodstad,V.Muller,K.Pantel,A.J.Minn,M.J.Bissell,B.A.Garcia,Y.Kang,V.K.Rajasekhar,C.M.Ghajar,I.Matei,H.Peinado,J.BrombergandD.Lyden(2015)."Tumourexosomeintegrinsdetermineorganotropicmetastasis."Nature527(7578):329-335.

Hu,W.,X.Zhang,Y.C.Tung,S.Xie,F.LiuandK.Iqbal(2016)."Hyperphosphorylationdeterminesboththespreadandthemorphologyoftaupathology."AlzheimersDement12(10):1066-1077.

Huhmer,A.F.,R.G.Biringer,H.Amato,A.N.FontehandM.G.Harrington(2006)."Proteinanalysisinhumancerebrospinalfluid:Physiologicalaspects,currentprogressandfuturechallenges."DisMarkers22(1-2):3-26.

Israel,M.A.,S.H.Yuan,C.Bardy,S.M.Reyna,Y.Mu,C.Herrera,M.P.Hefferan,S.VanGorp,K.L.Nazor,F.S.Boscolo,C.T.Carson,L.C.Laurent,M.Marsala,F.H.Gage,A.M.Remes,E.H.KooandL.S.Goldstein(2012)."ProbingsporadicandfamilialAlzheimer'sdiseaseusinginducedpluripotentstemcells."Nature482(7384):216-220.

109

Jiang,Y.,K.A.Mullaney,C.M.Peterhoff,S.Che,S.D.Schmidt,A.Boyer-Boiteau,S.D.Ginsberg,A.M.Cataldo,P.M.MathewsandR.A.Nixon(2010)."Alzheimer's-relatedendosomedysfunctioninDownsyndromeisAbeta-independentbutrequiresAPPandisreversedbyBACE-1inhibition."ProcNatlAcadSciUSA107(4):1630-1635.

Karch,C.M.andA.M.Goate(2015)."Alzheimer'sdiseaseriskgenesandmechanismsofdiseasepathogenesis."BiolPsychiatry77(1):43-51.

Karran,E.,M.MerckenandB.DeStrooper(2011)."TheamyloidcascadehypothesisforAlzheimer'sdisease:anappraisalforthedevelopmentoftherapeutics."NatRevDrugDiscov10(9):698-712.

Kerr,M.C.,J.S.Bennetts,F.Simpson,E.C.Thomas,C.Flegg,P.A.Gleeson,C.WickingandR.D.Teasdale(2005)."Anovelmammalianretromercomponent,Vps26B."Traffic6(11):991-1001.

Khalyfa,A.,D.GozalandL.Kheirandish-Gozal(2017)."PlasmaExosomesDisruptBloodBrainBarrierinChildrenwithOSAandNeurocognitiveDeficits."AmJRespirCritCareMed.

Khan,U.A.,L.Liu,F.A.Provenzano,D.E.Berman,C.P.Profaci,R.Sloan,R.Mayeux,K.E.DuffandS.A.Small(2014)."MoleculardriversandcorticalspreadoflateralentorhinalcortexdysfunctioninpreclinicalAlzheimer'sdisease."NatNeurosci17(2):304-311.

Kim,E.,Y.Lee,H.J.Lee,J.S.Kim,B.S.Song,J.W.Huh,S.R.Lee,S.U.Kim,S.H.Kim,Y.Hong,I.ShimandK.T.Chang(2010)."ImplicationofmouseVps26b-Vps29-Vps35retromercomplexinsortilintrafficking."BiochemBiophysResCommun403(2):167-171.

Kirschner,D.A.,H.Inouye,L.K.Duffy,A.Sinclair,M.LindandD.J.Selkoe(1987)."SyntheticpeptidehomologoustobetaproteinfromAlzheimerdiseaseformsamyloid-likefibrilsinvitro."ProcNatlAcadSciUSA84(19):6953-6957.

Klingeborn,M.,W.M.Dismuke,N.P.Skiba,U.Kelly,W.D.StamerandC.BowesRickman(2017)."DirectionalExosomeProteomesReflectPolarity-SpecificFunctionsinRetinalPigmentedEpitheliumMonolayers."SciRep7(1):4901.

KochanekKD,M.S.,XuJQ,Tejada-VeraB.(2016).Deaths:FinalDatafor2014.NationalVitalStatisticsReports.Hyattsville,Md.,NationalCenterforHealthStatistics.65.

Kosicek,M.,H.Zetterberg,N.Andreasen,J.Peter-KatalinicandS.Hecimovic(2012)."ElevatedcerebrospinalfluidsphingomyelinlevelsinprodromalAlzheimer'sdisease."NeurosciLett516(2):302-305.

Kosik,K.S.,C.L.JoachimandD.J.Selkoe(1986)."Microtubule-associatedproteintau(tau)isamajorantigeniccomponentofpairedhelicalfilamentsinAlzheimerdisease."ProcNatlAcadSciUSA83(11):4044-4048.

110

Kramer,A.,J.Green,J.Pollard,Jr.andS.Tugendreich(2014)."CausalanalysisapproachesinIngenuityPathwayAnalysis."Bioinformatics30(4):523-530.

LaFerla,F.M.,K.N.GreenandS.Oddo(2007)."Intracellularamyloid-betainAlzheimer'sdisease."NatRevNeurosci8(7):499-509.

Lambert,J.C.,C.A.Ibrahim-Verbaas,D.Harold,A.C.Naj,R.Sims,C.Bellenguez,A.L.DeStafano,J.C.Bis,G.W.Beecham,B.Grenier-Boley,G.Russo,T.A.Thorton-Wells,N.Jones,A.V.Smith,V.Chouraki,C.Thomas,M.A.Ikram,D.Zelenika,B.N.Vardarajan,Y.Kamatani,C.F.Lin,A.Gerrish,H.Schmidt,B.Kunkle,M.L.Dunstan,A.Ruiz,M.T.Bihoreau,S.H.Choi,C.Reitz,F.Pasquier,C.Cruchaga,D.Craig,N.Amin,C.Berr,O.L.Lopez,P.L.DeJager,V.Deramecourt,J.A.Johnston,D.Evans,S.Lovestone,L.Letenneur,F.J.Moron,D.C.Rubinsztein,G.Eiriksdottir,K.Sleegers,A.M.Goate,N.Fievet,M.W.Huentelman,M.Gill,K.Brown,M.I.Kamboh,L.Keller,P.Barberger-Gateau,B.McGuiness,E.B.Larson,R.Green,A.J.Myers,C.Dufouil,S.Todd,D.Wallon,S.Love,E.Rogaeva,J.Gallacher,P.StGeorge-Hyslop,J.Clarimon,A.Lleo,A.Bayer,D.W.Tsuang,L.Yu,M.Tsolaki,P.Bossu,G.Spalletta,P.Proitsi,J.Collinge,S.Sorbi,F.Sanchez-Garcia,N.C.Fox,J.Hardy,M.C.DenizNaranjo,P.Bosco,R.Clarke,C.Brayne,D.Galimberti,M.Mancuso,F.Matthews,I.EuropeanAlzheimer'sDisease,Genetic,D.EnvironmentalRiskinAlzheimer's,C.Alzheimer'sDiseaseGenetic,H.Cohortsfor,E.AgingResearchinGenomic,S.Moebus,P.Mecocci,M.DelZompo,W.Maier,H.Hampel,A.Pilotto,M.Bullido,F.Panza,P.Caffarra,B.Nacmias,J.R.Gilbert,M.Mayhaus,L.Lannefelt,H.Hakonarson,S.Pichler,M.M.Carrasquillo,M.Ingelsson,D.Beekly,V.Alvarez,F.Zou,O.Valladares,S.G.Younkin,E.Coto,K.L.Hamilton-Nelson,W.Gu,C.Razquin,P.Pastor,I.Mateo,M.J.Owen,K.M.Faber,P.V.Jonsson,O.Combarros,M.C.O'Donovan,L.B.Cantwell,H.Soininen,D.Blacker,S.Mead,T.H.Mosley,Jr.,D.A.Bennett,T.B.Harris,L.Fratiglioni,C.Holmes,R.F.deBruijn,P.Passmore,T.J.Montine,K.Bettens,J.I.Rotter,A.Brice,K.Morgan,T.M.Foroud,W.A.Kukull,D.Hannequin,J.F.Powell,M.A.Nalls,K.Ritchie,K.L.Lunetta,J.S.Kauwe,E.Boerwinkle,M.Riemenschneider,M.Boada,M.Hiltuenen,E.R.Martin,R.Schmidt,D.Rujescu,L.S.Wang,J.F.Dartigues,R.Mayeux,C.Tzourio,A.Hofman,M.M.Nothen,C.Graff,B.M.Psaty,L.Jones,J.L.Haines,P.A.Holmans,M.Lathrop,M.A.Pericak-Vance,L.J.Launer,L.A.Farrer,C.M.vanDuijn,C.VanBroeckhoven,V.Moskvina,S.Seshadri,J.Williams,G.D.SchellenbergandP.Amouyel(2013)."Meta-analysisof74,046individualsidentifies11newsusceptibilitylociforAlzheimer'sdisease."NatGenet45(12):1452-1458.

Lane,R.F.,S.M.Raines,J.W.Steele,M.E.Ehrlich,J.A.Lah,S.A.Small,R.E.Tanzi,A.D.AttieandS.Gandy(2010)."Diabetes-associatedSorCS1regulatesAlzheimer'samyloid-betametabolism:evidenceforinvolvementofSorL1andtheretromercomplex."JNeurosci30(39):13110-13115.

Lane,R.F.,P.StGeorge-Hyslop,B.L.Hempstead,S.A.Small,S.M.StrittmatterandS.Gandy(2012)."Vps10familyproteinsandtheretromercomplexinaging-relatedneurodegenerationanddiabetes."JNeurosci32(41):14080-14086.

111

Levy-Lahad,E.,W.Wasco,P.Poorkaj,D.M.Romano,J.Oshima,W.H.Pettingell,C.E.Yu,P.D.Jondro,S.D.Schmidt,K.Wangandetal.(1995)."Candidategeneforthechromosome1familialAlzheimer'sdiseaselocus."Science269(5226):973-977.

Luchsinger,J.A.(2008)."Adiposity,hyperinsulinemia,diabetesandAlzheimer'sdisease:anepidemiologicalperspective."EurJPharmacol585(1):119-129.

Luchsinger,J.A.,M.X.Tang,Y.Stern,S.SheaandR.Mayeux(2001)."DiabetesmellitusandriskofAlzheimer'sdiseaseanddementiawithstrokeinamultiethniccohort."AmJEpidemiol154(7):635-641.

Lucin,K.M.,C.E.O'Brien,G.Bieri,E.Czirr,K.I.Mosher,R.J.Abbey,D.F.Mastroeni,J.Rogers,B.Spencer,E.MasliahandT.Wyss-Coray(2013)."Microglialbeclin1regulatesretromertraffickingandphagocytosisandisimpairedinAlzheimer'sdisease."Neuron79(5):873-886.

Maia,L.F.,S.A.Kaeser,J.Reichwald,M.Hruscha,P.Martus,M.StaufenbielandM.Jucker(2013)."Changesinamyloid-betaandTauinthecerebrospinalfluidoftransgenicmiceoverexpressingamyloidprecursorprotein."SciTranslMed5(194):194re192.

Manczak,M.andP.H.Reddy(2013)."Abnormalinteractionofoligomericamyloid-betawithphosphorylatedtau:implicationstosynapticdysfunctionandneuronaldamage."JAlzheimersDis36(2):285-295.

Marquer,C.,H.Tian,J.Yi,J.Bastien,C.Dall'Armi,Y.Yang-Klingler,B.Zhou,R.B.ChanandG.DiPaolo(2016)."Arf6controlsretromertrafficandintracellularcholesteroldistributionviaaphosphoinositide-basedmechanism."NatCommun7:11919.

McNamara,M.J.,C.T.Ruff,W.Wasco,R.E.Tanzi,G.ThinakaranandB.T.Hyman(1998)."Immunohistochemicalandinsituanalysisofamyloidprecursor-likeprotein-1andamyloidprecursor-likeprotein-2expressioninAlzheimerdiseaseandagedcontrolbrains."BrainRes804(1):45-51.

Meredith,J.E.,Jr.,S.Sankaranarayanan,V.Guss,A.J.Lanzetti,F.Berisha,R.J.Neely,J.R.Slemmon,E.Portelius,H.Zetterberg,K.Blennow,H.Soares,M.AhlijanianandC.F.Albright(2013)."CharacterizationofnovelCSFTauandptaubiomarkersforAlzheimer'sdisease."PLoSOne8(10):e76523.

Michaelson,D.M.(2014)."APOEepsilon4:themostprevalentyetunderstudiedriskfactorforAlzheimer'sdisease."AlzheimersDement10(6):861-868.

Mim,C.andV.M.Unger(2012)."MembranecurvatureanditsgenerationbyBARproteins."TrendsBiochemSci37(12):526-533.

Miranda,A.M.,Z.M.Lasiecka,Y.Xu,J.Neufeld,S.Shahriar,S.Simoes,R.B.Chan,T.G.Oliveira,S.A.SmallandG.DiPaolo(2018)."NeuronallysosomaldysfunctionreleasesexosomesharboringAPPC-terminalfragmentsanduniquelipidsignatures."NatCommun9(1):291.

112

Morabito,M.V.,D.E.Berman,R.T.Schneider,Y.Zhang,R.L.LeibelandS.A.Small(2014)."HyperleucinemiacauseshippocampalretromerdeficiencylinkingdiabetestoAlzheimer'sdisease."NeurobiolDis65:188-192.

Morel,E.,Z.Chamoun,Z.M.Lasiecka,R.B.Chan,R.L.Williamson,C.Vetanovetz,C.Dall'Armi,S.Simoes,K.S.PointDuJour,B.D.McCabe,S.A.SmallandG.DiPaolo(2013)."Phosphatidylinositol-3-phosphateregulatessortingandprocessingofamyloidprecursorproteinthroughtheendosomalsystem."NatCommun4:2250.

Muhammad,A.,I.Flores,H.Zhang,R.Yu,A.Staniszewski,E.Planel,M.Herman,L.Ho,R.Kreber,L.S.Honig,B.Ganetzky,K.Duff,O.ArancioandS.A.Small(2008)."RetromerdeficiencyobservedinAlzheimer'sdiseasecauseshippocampaldysfunction,neurodegeneration,andAbetaaccumulation."ProcNatlAcadSciUSA105(20):7327-7332.

Nixon,R.A.(2017)."Amyloidprecursorproteinandendosomal-lysosomaldysfunctioninAlzheimer'sdisease:inseparablepartnersinamultifactorialdisease."FASEBJ31(7):2729-2743.

Novak,M.(1994)."TruncatedtauproteinasanewmarkerforAlzheimer'sdisease."ActaVirol38(3):173-189.

Nuriel,T.,K.Y.Peng,A.Ashok,A.A.Dillman,H.Y.Figueroa,J.Apuzzo,J.Ambat,E.Levy,M.R.Cookson,P.M.MathewsandK.E.Duff(2017)."TheEndosomal-LysosomalPathwayIsDysregulatedbyAPOE4ExpressioninVivo."FrontNeurosci11:702.

Offe,K.,S.E.Dodson,J.T.Shoemaker,J.J.Fritz,M.Gearing,A.I.LeveyandJ.J.Lah(2006)."ThelipoproteinreceptorLR11regulatesamyloidbetaproductionandamyloidprecursorproteintrafficinendosomalcompartments."JNeurosci26(5):1596-1603.

Perez-Gonzalez,R.,S.A.Gauthier,A.KumarandE.Levy(2012)."Theexosomesecretorypathwaytransportsamyloidprecursorproteincarboxyl-terminalfragmentsfromthecellintothebrainextracellularspace."JBiolChem287(51):43108-43115.

Pickford,F.,E.Masliah,M.Britschgi,K.Lucin,R.Narasimhan,P.A.Jaeger,S.Small,B.Spencer,E.Rockenstein,B.LevineandT.Wyss-Coray(2008)."Theautophagy-relatedproteinbeclin1showsreducedexpressioninearlyAlzheimerdiseaseandregulatesamyloidbetaaccumulationinmice."JClinInvest118(6):2190-2199.

Pottier,C.,D.Hannequin,S.Coutant,A.Rovelet-Lecrux,D.Wallon,S.Rousseau,S.Legallic,C.Paquet,S.Bombois,J.Pariente,C.Thomas-Anterion,A.Michon,B.Croisile,F.Etcharry-Bouyx,C.Berr,J.F.Dartigues,P.Amouyel,H.Dauchel,C.Boutoleau-Bretonniere,C.Thauvin,T.Frebourg,J.C.Lambert,D.CampionandP.G.Collaborators(2012)."HighfrequencyofpotentiallypathogenicSORL1mutationsinautosomaldominantearly-onsetAlzheimerdisease."MolPsychiatry17(9):875-879.

113

Priller,C.,T.Bauer,G.Mitteregger,B.Krebs,H.A.KretzschmarandJ.Herms(2006)."Synapseformationandfunctionismodulatedbytheamyloidprecursorprotein."JNeurosci26(27):7212-7221.

Raja,W.K.,A.E.Mungenast,Y.T.Lin,T.Ko,F.Abdurrob,J.SeoandL.H.Tsai(2016)."Self-Organizing3DHumanNeuralTissueDerivedfromInducedPluripotentStemCellsRecapitulateAlzheimer'sDiseasePhenotypes."PLoSOne11(9):e0161969.

Reinhardt,T.A.,R.E.Sacco,B.J.NonneckeandJ.D.Lippolis(2013)."Bovinemilkproteome:quantitativechangesinnormalmilkexosomes,milkfatglobulemembranesandwheyproteomesresultingfromStaphylococcusaureusmastitis."JProteomics82:141-154.

Rentz,D.M.,J.J.Locascio,J.A.Becker,E.K.Moran,E.Eng,R.L.Buckner,R.A.SperlingandK.A.Johnson(2010)."Cognition,reserve,andamyloiddepositioninnormalaging."AnnNeurol67(3):353-364.

Rodriguez,L.,N.V.Mohamed,A.Desjardins,R.Lippe,E.A.FonandN.Leclerc(2017)."Rab7Aregulatestausecretion."JNeurochem141(4):592-605.

Rogaeva,E.,Y.Meng,J.H.Lee,Y.Gu,T.Kawarai,F.Zou,T.Katayama,C.T.Baldwin,R.Cheng,H.Hasegawa,F.Chen,N.Shibata,K.L.Lunetta,R.Pardossi-Piquard,C.Bohm,Y.Wakutani,L.A.Cupples,K.T.Cuenco,R.C.Green,L.Pinessi,I.Rainero,S.Sorbi,A.Bruni,R.Duara,R.P.Friedland,R.Inzelberg,W.Hampe,H.Bujo,Y.Q.Song,O.M.Andersen,T.E.Willnow,N.Graff-Radford,R.C.Petersen,D.Dickson,S.D.Der,P.E.Fraser,G.Schmitt-Ulms,S.Younkin,R.Mayeux,L.A.FarrerandP.StGeorge-Hyslop(2007)."Theneuronalsortilin-relatedreceptorSORL1isgeneticallyassociatedwithAlzheimerdisease."NatGenet39(2):168-177.

Rojas,R.,S.Kametaka,C.R.HaftandJ.S.Bonifacino(2007)."InterchangeablebutessentialfunctionsofSNX1andSNX2intheassociationofretromerwithendosomesandthetraffickingofmannose6-phosphatereceptors."MolCellBiol27(3):1112-1124.

Rojas,R.,T.vanVlijmen,G.A.Mardones,Y.Prabhu,A.L.Rojas,S.Mohammed,A.J.Heck,G.Raposo,P.vanderSluijsandJ.S.Bonifacino(2008)."RegulationofretromerrecruitmenttoendosomesbysequentialactionofRab5andRab7."JCellBiol183(3):513-526.

Romeo,G.R.,K.S.MoultonandA.Kazlauskas(2007)."Attenuatedexpressionofprofilin-1confersprotectionfromatherosclerosisintheLDLreceptornullmouse."CircRes101(4):357-367.

Rozek,W.,M.Ricardo-Dukelow,S.Holloway,H.E.Gendelman,V.Wojna,L.M.MelendezandP.Ciborowski(2007)."CerebrospinalfluidproteomicprofilingofHIV-1-infectedpatientswithcognitiveimpairment."JProteomeRes6(11):4189-4199.

Saman,S.,W.Kim,M.Raya,Y.Visnick,S.Miro,S.Saman,B.Jackson,A.C.McKee,V.E.Alvarez,N.C.LeeandG.F.Hall(2012)."Exosome-associatedtauissecretedintauopathymodelsandis

114

selectivelyphosphorylatedincerebrospinalfluidinearlyAlzheimerdisease."JBiolChem287(6):3842-3849.

Scarmeas,N.,J.A.Luchsinger,N.Schupf,A.M.Brickman,S.Cosentino,M.X.TangandY.Stern(2009)."Physicalactivity,diet,andriskofAlzheimerdisease."JAMA302(6):627-637.

Scarmeas,N.,Y.Stern,M.X.Tang,R.MayeuxandJ.A.Luchsinger(2006)."MediterraneandietandriskforAlzheimer'sdisease."AnnNeurol59(6):912-921.

Scarmeas,N.,E.Zarahn,K.E.Anderson,C.G.Habeck,J.Hilton,J.Flynn,K.S.Marder,K.L.Bell,H.A.Sackeim,R.L.VanHeertum,J.R.MoellerandY.Stern(2003)."AssociationoflifeactivitieswithcerebralbloodflowinAlzheimerdisease:implicationsforthecognitivereservehypothesis."ArchNeurol60(3):359-365.

Scarmeas,N.,E.Zarahn,K.E.Anderson,L.S.Honig,A.Park,J.Hilton,J.Flynn,H.A.SackeimandY.Stern(2004)."Cognitivereserve-mediatedmodulationofpositronemissiontomographicactivationsduringmemorytasksinAlzheimerdisease."ArchNeurol61(1):73-78.

Schelle,J.,L.M.Hasler,J.C.Gopfert,T.O.Joos,H.Vanderstichele,E.Stoops,E.M.Mandelkow,U.Neumann,D.R.Shimshek,M.Staufenbiel,M.JuckerandS.A.Kaeser(2017)."PreventionoftauincreaseincerebrospinalfluidofAPPtransgenicmicesuggestsdownstreameffectofBACE1inhibition."AlzheimersDement13(6):701-709.

Schutzer,S.E.,T.Liu,B.H.Natelson,T.E.Angel,A.A.Schepmoes,S.O.Purvine,K.K.Hixson,M.S.Lipton,D.G.Camp,P.K.Coyle,R.D.SmithandJ.Bergquist(2010)."Establishingtheproteomeofnormalhumancerebrospinalfluid."PLoSOne5(6):e10980.

Schwagerl,A.L.,P.S.Mohan,A.M.Cataldo,J.P.Vonsattel,N.W.KowallandR.A.Nixon(1995)."Elevatedlevelsoftheendosomal-lysosomalproteinasecathepsinDincerebrospinalfluidinAlzheimerdisease."JNeurochem64(1):443-446.

Seaman,M.N.(2004)."Cargo-selectiveendosomalsortingforretrievaltotheGolgirequiresretromer."JCellBiol165(1):111-122.

Seaman,M.N.,M.E.Harbour,D.Tattersall,E.ReadandN.Bright(2009)."Membranerecruitmentofthecargo-selectiveretromersubcomplexiscatalysedbythesmallGTPaseRab7andinhibitedbytheRab-GAPTBC1D5."JCellSci122(Pt14):2371-2382.

Seaman,M.N.,E.G.Marcusson,J.L.CereghinoandS.D.Emr(1997)."EndosometoGolgiretrievalofthevacuolarproteinsortingreceptor,Vps10p,requiresthefunctionoftheVPS29,VPS30,andVPS35geneproducts."JCellBiol137(1):79-92.

Seaman,M.N.,J.M.McCafferyandS.D.Emr(1998)."Amembranecoatcomplexessentialforendosome-to-Golgiretrogradetransportinyeast."JCellBiol142(3):665-681.

115

Serrano-Pozo,A.,M.P.Frosch,E.MasliahandB.T.Hyman(2011)."NeuropathologicalalterationsinAlzheimerdisease."ColdSpringHarbPerspectMed1(1):a006189.

Seyer,A.,S.Boudah,S.Broudin,C.JunotandB.Colsch(2016)."Annotationofthehumancerebrospinalfluidlipidomeusinghighresolutionmassspectrometryandadedicateddataprocessingworkflow."Metabolomics12:91.

Sherrington,R.,E.I.Rogaev,Y.Liang,E.A.Rogaeva,G.Levesque,M.Ikeda,H.Chi,C.Lin,G.Li,K.Holman,T.Tsuda,L.Mar,J.F.Foncin,A.C.Bruni,M.P.Montesi,S.Sorbi,I.Rainero,L.Pinessi,L.Nee,I.Chumakov,D.Pollen,A.Brookes,P.Sanseau,R.J.Polinsky,W.Wasco,H.A.DaSilva,J.L.Haines,M.A.Perkicak-Vance,R.E.Tanzi,A.D.Roses,P.E.Fraser,J.M.RommensandP.H.StGeorge-Hyslop(1995)."Cloningofagenebearingmissensemutationsinearly-onsetfamilialAlzheimer'sdisease."Nature375(6534):754-760.

Shibata,M.,S.Yamada,S.R.Kumar,M.Calero,J.Bading,B.Frangione,D.M.Holtzman,C.A.Miller,D.K.Strickland,J.GhisoandB.V.Zlokovic(2000)."ClearanceofAlzheimer'samyloid-ss(1-40)peptidefrombrainbyLDLreceptor-relatedprotein-1attheblood-brainbarrier."JClinInvest106(12):1489-1499.

Shores,K.S.andD.R.Knapp(2007)."Assessmentapproachforevaluatinghighabundanceproteindepletionmethodsforcerebrospinalfluid(CSF)proteomicanalysis."JProteomeRes6(9):3739-3751.

Skogberg,G.,J.Gudmundsdottir,S.vanderPost,K.Sandstrom,S.Bruhn,M.Benson,L.Mincheva-Nilsson,V.Baranov,E.TelemoandO.Ekwall(2013)."Characterizationofhumanthymicexosomes."PLoSOne8(7):e67554.

Small,S.A.andK.Duff(2008)."LinkingAbetaandtauinlate-onsetAlzheimer'sdisease:adualpathwayhypothesis."Neuron60(4):534-542.

Small,S.A.andS.Gandy(2006)."SortingthroughthecellbiologyofAlzheimer'sdisease:intracellularpathwaystopathogenesis."Neuron52(1):15-31.

Small,S.A.,K.Kent,A.Pierce,C.Leung,M.S.Kang,H.Okada,L.Honig,J.P.VonsattelandT.W.Kim(2005)."Model-guidedmicroarrayimplicatestheretromercomplexinAlzheimer'sdisease."AnnNeurol58(6):909-919.

Small,S.A.andG.A.Petsko(2015)."RetromerinAlzheimerdisease,Parkinsondiseaseandotherneurologicaldisorders."NatRevNeurosci16(3):126-132.

Small,S.A.,S.Simoes-Spassov,R.MayeuxandG.A.Petsko(2017)."EndosomalTrafficJamsRepresentaPathogenicHubandTherapeuticTargetinAlzheimer'sDisease."TrendsNeurosci40(10):592-602.

116

Smith,J.S.,T.E.Angel,C.Chavkin,D.J.Orton,R.J.MooreandR.D.Smith(2014)."Characterizationofindividualmousecerebrospinalfluidproteomes."Proteomics14(9):1102-1106.

Steinberg,S.,H.Stefansson,T.Jonsson,H.Johannsdottir,A.Ingason,H.Helgason,P.Sulem,O.T.Magnusson,S.A.Gudjonsson,U.Unnsteinsdottir,A.Kong,S.Helisalmi,H.Soininen,J.J.Lah,DemGene,D.Aarsland,T.Fladby,I.D.Ulstein,S.Djurovic,S.B.Sando,L.R.White,G.P.Knudsen,L.T.Westlye,G.Selbaek,I.Giegling,H.Hampel,M.Hiltunen,A.I.Levey,O.A.Andreassen,D.Rujescu,P.V.Jonsson,S.Bjornsson,J.SnaedalandK.Stefansson(2015)."Loss-of-functionvariantsinABCA7conferriskofAlzheimer'sdisease."NatGenet47(5):445-447.

Stern,Y.(2006)."CognitivereserveandAlzheimerdisease."AlzheimerDisAssocDisord20(2):112-117.

Strauss,K.,C.Goebel,H.Runz,W.Mobius,S.Weiss,I.Feussner,M.SimonsandA.Schneider(2010)."ExosomesecretionameliorateslysosomalstorageofcholesterolinNiemann-PicktypeCdisease."JBiolChem285(34):26279-26288.

Strittmatter,W.J.,K.H.Weisgraber,D.Y.Huang,L.M.Dong,G.S.Salvesen,M.Pericak-Vance,D.Schmechel,A.M.Saunders,D.GoldgaberandA.D.Roses(1993)."BindingofhumanapolipoproteinEtosyntheticamyloidbetapeptide:isoform-specificeffectsandimplicationsforlate-onsetAlzheimerdisease."ProcNatlAcadSciUSA90(17):8098-8102.

Sullivan,C.P.,A.G.Jay,E.C.Stack,M.Pakaluk,E.Wadlinger,R.E.Fine,J.M.WellsandP.J.Morin(2011)."RetromerdisruptionpromotesamyloidogenicAPPprocessing."NeurobiolDis43(2):338-345.

Tanzi,R.E.,R.D.MoirandS.L.Wagner(2004)."ClearanceofAlzheimer'sAbetapeptide:themanyroadstoperdition."Neuron43(5):605-608.

Teller,J.K.,C.Russo,L.M.DeBusk,G.Angelini,D.Zaccheo,F.Dagna-Bricarelli,P.Scartezzini,S.Bertolini,D.M.Mann,M.TabatonandP.Gambetti(1996)."Presenceofsolubleamyloidbeta-peptideprecedesamyloidplaqueformationinDown'ssyndrome."NatMed2(1):93-95.

Temkin,P.,W.Morishita,D.Goswami,K.Arendt,L.ChenandR.Malenka(2017)."TheRetromerSupportsAMPAReceptorTraffickingDuringLTP."Neuron94(1):74-82e75.

Tholen,S.,M.L.Biniossek,M.Gansz,T.D.Ahrens,M.Schlimpert,J.N.Kizhakkedathu,T.ReinheckelandO.Schilling(2014)."DoubledeficiencyofcathepsinsBandLresultsinmassivesecretomealterationsandsuggestsadegradativecathepsin-MMPaxis."CellMolLifeSci71(5):899-916.

Tholen,S.,M.L.Biniossek,A.L.Gessler,S.Muller,J.Weisser,J.N.Kizhakkedathu,T.ReinheckelandO.Schilling(2011)."ContributionofcathepsinLtosecretomecompositionandcleavagepatternofmouseembryonicfibroblasts."BiolChem392(11):961-971.

117

Trajkovic,K.,C.Hsu,S.Chiantia,L.Rajendran,D.Wenzel,F.Wieland,P.Schwille,B.BruggerandM.Simons(2008)."Ceramidetriggersbuddingofexosomevesiclesintomultivesicularendosomes."Science319(5867):1244-1247.

Tran,J.,M.Mustapic,N.Voigt,B.BendlinandD.Kapogiannis2017NeuroscienceMeetingPlanner.Washingto,D.C.SocietyforNeuroscience,2017.Online.

.

Tran,J.M.,M.;Voigt,N.;Bendlin,B.;Kapogiannis,D.(2017).ExtracellularVesicleBiomarkersinAlzheimer’sDisease,MildCognitiveImpairment,andNon-CognitivelyImpairedOlderAdults.2017NeuroscienceMeetingPlanner,Washingto,D.C.,SocietyforNeuroscience,2017.

Tyan,S.H.,A.Y.Shih,J.J.Walsh,H.Maruyama,F.Sarsoza,L.Ku,S.Eggert,P.R.Hof,E.H.KooandD.L.Dickstein(2012)."Amyloidprecursorprotein(APP)regulatessynapticstructureandfunction."MolCellNeurosci51(1-2):43-52.

Vardarajan,B.N.,S.Y.Bruesegem,M.E.Harbour,R.Inzelberg,R.Friedland,P.StGeorge-Hyslop,M.N.SeamanandL.A.Farrer(2012)."IdentificationofAlzheimerdisease-associatedvariantsingenesthatregulateretromerfunction."NeurobiolAging33(9):2231e2215-2231e2230.

Vardarajan,B.N.,Y.Zhang,J.H.Lee,R.Cheng,C.Bohm,M.Ghani,C.Reitz,D.Reyes-Dumeyer,Y.Shen,E.Rogaeva,P.StGeorge-HyslopandR.Mayeux(2015)."CodingmutationsinSORL1andAlzheimerdisease."AnnNeurol77(2):215-227.

Verheijen,J.,T.VandenBossche,J.vanderZee,S.Engelborghs,R.Sanchez-Valle,A.Llado,C.Graff,H.Thonberg,P.Pastor,S.Ortega-Cubero,M.A.Pastor,L.Benussi,R.Ghidoni,G.Binetti,J.Clarimon,A.Lleo,J.Fortea,A.deMendonca,M.Martins,O.Grau-Rivera,E.Gelpi,K.Bettens,L.Mateiu,L.Dillen,P.Cras,P.P.DeDeyn,C.VanBroeckhovenandK.Sleegers(2016)."AcomprehensivestudyofthegeneticimpactofrarevariantsinSORL1inEuropeanearly-onsetAlzheimer'sdisease."ActaNeuropathol132(2):213-224.

Walsh,D.M.,I.Klyubin,J.V.Fadeeva,W.K.Cullen,R.Anwyl,M.S.Wolfe,M.J.RowanandD.J.Selkoe(2002)."Naturallysecretedoligomersofamyloidbetaproteinpotentlyinhibithippocampallong-termpotentiationinvivo."Nature416(6880):535-539.

Wang,J.Z.,Y.Y.Xia,I.Grundke-IqbalandK.Iqbal(2013)."Abnormalhyperphosphorylationoftau:sites,regulation,andmolecularmechanismofneurofibrillarydegeneration."JAlzheimersDis33Suppl1:S123-139.

Wang,Y.,V.Balaji,S.Kaniyappan,L.Kruger,S.Irsen,K.Tepper,R.Chandupatla,W.Maetzler,A.Schneider,E.MandelkowandE.M.Mandelkow(2017)."Thereleaseandtrans-synaptictransmissionofTauviaexosomes."MolNeurodegener12(1):5.

118

Wassmer,T.,N.Attar,M.V.Bujny,J.Oakley,C.J.TraerandP.J.Cullen(2007)."Aloss-of-functionscreenrevealsSNX5andSNX6aspotentialcomponentsofthemammalianretromer."JCellSci120(Pt1):45-54.

Wen,L.,F.L.Tang,Y.Hong,S.W.Luo,C.L.Wang,W.He,C.Shen,J.U.Jung,F.Xiong,D.H.Lee,Q.G.Zhang,D.Brann,T.W.Kim,R.Yan,L.MeiandW.C.Xiong(2011)."VPS35haploinsufficiencyincreasesAlzheimer'sdiseaseneuropathology."JCellBiol195(5):765-779.

Wibowo,A.S.,M.Singh,K.M.Reeder,J.J.Carter,A.R.Kovach,W.Meng,M.Ratnam,F.ZhangandC.E.Dann,3rd(2013)."Structuresofhumanfolatereceptorsrevealbiologicaltraffickingstatesanddiversityinfolateandantifolaterecognition."ProcNatlAcadSciUSA110(38):15180-15188.

Wu,D.,M.D.MilutinovicandD.R.Walt(2015)."Singlemoleculearray(Simoa)assaywithoptimalantibodypairsforcytokinedetectioninhumanserumsamples."Analyst140(18):6277-6282.

Wu,L.andL.Zhao(2016)."ApoE2andAlzheimer'sdisease:timetotakeacloserlook."NeuralRegenRes11(3):412-413.

Xiao,T.,W.Zhang,B.Jiao,C.Z.Pan,X.LiuandL.Shen(2017)."TheroleofexosomesinthepathogenesisofAlzheimer'disease."TranslNeurodegener6:3.

Xiong,H.,D.Callaghan,A.Jones,D.G.Walker,L.F.Lue,T.G.Beach,L.I.Sue,J.Woulfe,H.Xu,D.B.StanimirovicandW.Zhang(2008)."CholesterolretentioninAlzheimer'sbrainisresponsibleforhighbeta-andgamma-secretaseactivitiesandAbetaproduction."NeurobiolDis29(3):422-437.

Xu,W.,A.M.Weissmiller,J.A.White,2nd,F.Fang,X.Wang,Y.Wu,M.L.Pearn,X.Zhao,M.Sawa,S.Chen,S.Gunawardena,J.Ding,W.C.MobleyandC.Wu(2016)."Amyloidprecursorprotein-mediatedendocyticpathwaydisruptioninducesaxonaldysfunctionandneurodegeneration."JClinInvest126(5):1815-1833.

Yang,T.,P.Martin,B.Fogarty,A.Brown,K.Schurman,R.Phipps,V.P.Yin,P.LockmanandS.Bai(2015)."Exosomedeliveredanticancerdrugsacrosstheblood-brainbarrierforbraincancertherapyinDaniorerio."PharmRes32(6):2003-2014.

Young,J.,L.Fong,H.Frankowski,G.Petsko,S.A.SmallandL.Goldstein(2018)."StabilizingtheretromercomplexinahumanstemcellmodelofAlzheimer’sdiseasereducesTauphosphorylationindependentofAmyloidPrecursorProtein."StemCellReportsInPress.

Yuan,X.andD.M.Desiderio(2005)."Proteomicsanalysisofprefractionatedhumanlumbarcerebrospinalfluid."Proteomics5(2):541-550.

119

Yuyama,K.,N.YamamotoandK.Yanagisawa(2008)."Acceleratedreleaseofexosome-associatedGM1ganglioside(GM1)byendocyticpathwayabnormality:anotherputativepathwayforGM1-inducedamyloidfibrilformation."JNeurochem105(1):217-224.

Zarrouk,A.,M.Debbabi,M.Bezine,E.M.Karym,A.Badreddine,O.Rouaud,T.Moreau,M.Cherkaoui-Malki,M.ElAyeb,B.Nasser,M.HammamiandG.Lizard(2017)."LipidBiomarkersinAlzheimer'sDisease."CurrAlzheimerRes.

Zetterberg,H.(2017)."Review:Tauinbiofluids-relationtopathology,imagingandclinicalfeatures."NeuropatholApplNeurobiol43(3):194-199.

Zhang,L.,S.Zhang,J.Yao,F.J.Lowery,Q.Zhang,W.C.Huang,P.Li,M.Li,X.Wang,C.Zhang,H.Wang,K.Ellis,M.Cheerathodi,J.H.McCarty,D.Palmieri,J.Saunus,S.Lakhani,S.Huang,A.A.Sahin,K.D.Aldape,P.S.SteegandD.Yu(2015)."Microenvironment-inducedPTENlossbyexosomalmicroRNAprimesbrainmetastasisoutgrowth."Nature527(7576):100-104.

Zhang,Y.,K.Chen,S.A.Sloan,M.L.Bennett,A.R.Scholze,S.O'Keeffe,H.P.Phatnani,P.Guarnieri,C.Caneda,N.Ruderisch,S.Deng,S.A.Liddelow,C.Zhang,R.Daneman,T.Maniatis,B.A.BarresandJ.Q.Wu(2014)."AnRNA-sequencingtranscriptomeandsplicingdatabaseofglia,neurons,andvascularcellsofthecerebralcortex."JNeurosci34(36):11929-11947.

Zhao,X.,S.Nothwehr,R.Lara-Lemus,B.Y.Zhang,H.PeterandP.Arvan(2007)."Dominant-negativebehaviorofmammalianVps35inyeastrequiresaconservedPRLYLmotifinvolvedinretromerassembly."Traffic8(12):1829-1840.

Zheng,W.H.,S.Bastianetto,F.Mennicken,W.MaandS.Kar(2002)."Amyloidbetapeptideinducestauphosphorylationandlossofcholinergicneuronsinratprimaryseptalcultures."Neuroscience115(1):201-211.

Zlokovic,B.V.,C.L.Martel,E.Matsubara,J.G.McComb,G.Zheng,R.T.McCluskey,B.FrangioneandJ.Ghiso(1996)."Glycoprotein330/megalin:probableroleinreceptor-mediatedtransportofapolipoproteinJaloneandinacomplexwithAlzheimerdiseaseamyloidbetaattheblood-brainandblood-cerebrospinalfluidbarriers."ProcNatlAcadSciUSA93(9):4229-4234.